<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:wfw="http://wellformedweb.org/CommentAPI/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	xmlns:slash="http://purl.org/rss/1.0/modules/slash/"
	 xmlns:media="http://search.yahoo.com/mrss/" xmlns:itunes="http://www.itunes.com/dtds/podcast-1.0.dtd" xmlns:media="http://search.yahoo.com/mrss/"
>

<channel>
	<title>RPABOTS.WORLD</title>
	<atom:link href="https://rpabotsworld.com/tag/mcp/feed/" rel="self" type="application/rss+xml" />
	<link>https://rpabotsworld.com</link>
	<description>RPA, Agentic AI &amp; Intelligent Automation — Tutorials, Tools &amp; Career Guides</description>
	<lastBuildDate>Fri, 30 Oct 2026 03:10:49 +0000</lastBuildDate>
	<language>en-US</language>
	<sy:updatePeriod>
	hourly	</sy:updatePeriod>
	<sy:updateFrequency>
	1	</sy:updateFrequency>
	<generator>https://wordpress.org/?v=7.1.5</generator>
	<itunes:subtitle>RPABOTS.WORLD</itunes:subtitle>
	<itunes:summary>RPA, Agentic AI &amp; Intelligent Automation — Tutorials, Tools &amp; Career Guides</itunes:summary>
	<itunes:explicit>clean</itunes:explicit>
	<item>
		<title>Tenable CyberAgents Exchange AI Inspector: The Agentic AI Architect&#8217;s Security Guide (2026)</title>
		<link>https://rpabotsworld.com/tenable-cyberagents-exchange-ai-inspector-agent-security-guide/</link>
					<comments>https://rpabotsworld.com/tenable-cyberagents-exchange-ai-inspector-agent-security-guide/#respond</comments>
		
		<dc:creator><![CDATA[Satish Prasad]]></dc:creator>
		<pubDate>Mon, 26 Oct 2026 05:10:43 +0000</pubDate>
				<category><![CDATA[AI Agents & Frameworks]]></category>
		<category><![CDATA[RPA & Bot Automation]]></category>
		<category><![CDATA[MCP]]></category>
		<guid isPermaLink="false">https://rpabotsworld.com/?p=32367</guid>

					<description><![CDATA[Tenable's CyberAgents Exchange AI Inspector combines OpenAI GPT cyber models with expert review to vet AI agents, MCP servers, and playbooks before deployment.]]></description>
										<content:encoded><![CDATA[<h2>Why Your AI Agents Need a Security Gate — and Who Just Built One</h2>
<p>A Dark Reading survey found that nearly half of cybersecurity professionals view agentic AI as the leading attack vector heading into 2026. Meanwhile, the <a href="https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai" target="_blank" rel="noopener nofollow">McKinsey State of AI 2026</a> report shows 40% of large enterprises are now scaling AI agents in at least one function — up from 27% last year. The gap between adoption velocity and security readiness is widening, not narrowing.</p>
<p>On September 3, 2026, Tenable and OpenAI moved to close that gap. At OpenAI&#8217;s Intelligence at Work: Cyber Summit, Tenable unveiled the <strong>CyberAgents Exchange AI Inspector</strong> — the first purpose-built security review pipeline that combines frontier AI assessment with human expert review to inspect AI agents, skills, MCP servers, and multi-agent playbooks before they reach production environments.</p>
<p>For Agentic AI Architects, this is not just another vendor announcement. It is the emergence of a security primitive that the agent ecosystem has been missing: a structured, repeatable way to vet the components you are assembling into enterprise agent systems. This guide breaks down what the CyberAgents Exchange is, how the AI Inspector works, why it matters for your architecture decisions, and how to start using it today.</p>
<h2>Table of Contents</h2>
<ul>
<li><a href="#what-is-cyberagents-exchange">What Is the CyberAgents Exchange?</a></li>
<li><a href="#ai-inspector-architecture">How the AI Inspector Works: Three-Layer Security Review</a></li>
<li><a href="#why-agent-security-matters">The Agent Security Problem in 2026</a></li>
<li><a href="#openai-daybreak-context">The OpenAI Daybreak Defense Network Context</a></li>
<li><a href="#component-types">What Gets Inspected: Agents, Skills, MCP Servers, and Playbooks</a></li>
<li><a href="#for-architects">What This Means for Agentic AI Architects</a></li>
<li><a href="#getting-started">Getting Started: Contributing and Consuming</a></li>
<li><a href="#comparison">How CyberAgents Exchange Compares to Other Agent Registries</a></li>
<li><a href="#faqs">FAQs</a></li>
<li><a href="#key-takeaways">Key Takeaways</a></li>
<li><a href="#references">References</a></li>
</ul>
<h2 id="what-is-cyberagents-exchange">What Is the CyberAgents Exchange?</h2>
<p>The <a href="https://exchange.tenable.com/" target="_blank" rel="noopener nofollow">CyberAgents Exchange</a> is an open-source, vendor-agnostic registry for cybersecurity AI components. Tenable launched it on August 6, 2026 with over 50 initial components. By the time the AI Inspector was announced a month later, the registry had grown to more than 100 community-submitted entries, boosted by Tenable&#8217;s SWARM build event at Black Hat USA 2026.</p>
<p>The Exchange is not a code host — it is a curated directory. Each listing is a single Markdown file with YAML frontmatter that holds metadata (author, version, license, component type, tags), while the actual agent code lives in the contributor&#8217;s own GitHub repository. This design mirrors how package registries like npm or PyPI work: the registry indexes and vets, but the source of truth stays with the maintainer.</p>
<p>The <a href="https://github.com/tenable/cyberagents-exchange" target="_blank" rel="noopener nofollow">GitHub repository (tenable/cyberagents-exchange)</a> is organized into four directories matching the four component types the Exchange indexes:</p>
<table>
<thead>
<tr>
<th>Component Type</th>
<th>What It Is</th>
<th>Example</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Agents</strong></td>
<td>Autonomous AI tools that perform security tasks end-to-end</td>
<td>Navi — CLI tool integrating with Tenable One Vulnerability Management APIs</td>
</tr>
<tr>
<td><strong>Skills</strong></td>
<td>Reusable capabilities that extend AI coding assistants with security functions</td>
<td>CVE lookup skill, SBOM analysis skill</td>
</tr>
<tr>
<td><strong>MCP Servers</strong></td>
<td>Model Context Protocol servers that connect AI systems with security tools and data sources</td>
<td>SIEM connector, threat feed MCP server</td>
</tr>
<tr>
<td><strong>Playbooks</strong></td>
<td>Multi-agent workflows chaining several agents into end-to-end security processes</td>
<td>Incident triage → enrichment → remediation pipeline</td>
</tr>
</tbody>
</table>
<p>Founding members <strong>SentinelOne</strong> and <strong>Recorded Future</strong> joined at launch, contributing their expertise in autonomous security operations and threat intelligence respectively. The Exchange is free to use and contribute to — no Tenable product license is required.</p>
<h2 id="ai-inspector-architecture">How the AI Inspector Works: Three-Layer Security Review</h2>
<p>The CyberAgents Exchange AI Inspector, announced at the OpenAI Cyber Summit on September 3, 2026, is not a single tool — it is a three-layer inspection pipeline that combines automated AI analysis with human expert judgment. Here is how each layer works.</p>
<h3>Layer 1: Frontier Assessment Using OpenAI GPT Cyber Models</h3>
<p>The first layer uses OpenAI&#8217;s specialized GPT cyber models — the same class of models available to partners in the <a href="https://openai.com/daybreak/" target="_blank" rel="noopener nofollow">Daybreak Defense Network</a>. These models perform automated code analysis on submitted components, checking for:</p>
<ul>
<li><strong>Prompt injection vulnerabilities</strong> — both direct (malicious user input) and indirect (hidden instructions in data the agent retrieves)</li>
<li><strong>Tool misuse patterns</strong> — agent actions that could escalate privileges, exfiltrate data, or chain into unintended side effects</li>
<li><strong>Dependency and supply chain risks</strong> — flagging known-vulnerable libraries, suspicious imports, or unsigned dependencies</li>
<li><strong>Overpermissioned configurations</strong> — agents requesting more access than their stated function requires</li>
</ul>
<p>The frontier AI layer can analyze components at a depth and speed that would be impractical for human reviewers alone — scanning entire MCP server codebases, tracing data flows through multi-agent playbooks, and identifying subtle interaction patterns across chained agents.</p>
<h3>Layer 2: Skills Inspection via Tenable One AI Exposure</h3>
<p>The second layer feeds the component through <a href="https://www.tenable.com/products/ai-exposure" target="_blank" rel="noopener nofollow">Tenable One AI Exposure</a>, Tenable&#8217;s production security platform. This layer evaluates:</p>
<ul>
<li><strong>Attack surface exposure</strong> — what external interfaces the component exposes and how they could be exploited in a real enterprise environment</li>
<li><strong>Configuration drift detection</strong> — whether the component&#8217;s security posture would degrade when deployed alongside typical enterprise tooling</li>
<li><strong>Identity and access patterns</strong> — how the component authenticates, what non-human identities it creates, and whether those identities follow least-privilege principles</li>
</ul>
<p>This is where the Inspector moves from &#8220;is this code safe in isolation?&#8221; to &#8220;is this code safe in your environment?&#8221; — a critical distinction that pure static analysis misses.</p>
<h3>Layer 3: Expert Review by Tenable Researchers</h3>
<p>The third layer is human. Tenable&#8217;s security research team manually reviews flagged components, assessing context-dependent risks that automated systems cannot fully evaluate: social engineering vectors in agent conversation flows, domain-specific misuse potential, and architectural patterns that look benign in isolation but create emergent risks when composed with other agents.</p>
<p>This three-layer approach — automated AI scan, production security platform analysis, human expert review — is what distinguishes the AI Inspector from simpler approaches like running a linter or a static analysis tool. It mirrors the defense-in-depth model that security teams already use for software supply chains, applied to the new category of AI agent components.</p>
<h2 id="why-agent-security-matters">The Agent Security Problem in 2026</h2>
<p>The CyberAgents Exchange AI Inspector did not emerge in a vacuum. It is a direct response to a set of compounding security risks that have intensified throughout 2026 as agent adoption has accelerated. Understanding these risks is essential context for any architect deciding whether — and how — to integrate the Inspector into their workflow.</p>
<h3>The Adoption-Security Gap</h3>
<p>The numbers tell a stark story. According to <a href="https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai" target="_blank" rel="noopener nofollow">McKinsey&#8217;s State of AI 2026</a>, 62% of organizations are at least experimenting with AI agents, with 23% scaling in at least one function. But only 14.4% of agents that went live had full security and IT approval, according to the Gravitee State of AI Agent Security 2026 report. And <a href="https://aembit.io/blog/agentic-ai-cybersecurity-risks-security-guide/" target="_blank" rel="noopener nofollow">Okta&#8217;s survey of 260 executives</a> found that only 10% of organizations have a well-developed strategy for managing non-human and agentic identities.</p>
<p>This means the vast majority of production agents are running with improvised security, not engineered security. The components being assembled — agents, skills, MCP servers, playbooks — are community-built, open-source, and largely unvetted.</p>
<h3>The Five Attack Vectors</h3>
<p>Security researchers and practitioners have identified five primary attack categories targeting agentic AI systems in 2026:</p>
<table>
<thead>
<tr>
<th>Attack Vector</th>
<th>What Happens</th>
<th>Why Agents Are Especially Vulnerable</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Prompt Injection</strong></td>
<td>Adversary injects instructions that alter agent behavior</td>
<td>Agents process external data (emails, documents, web pages) where hidden instructions can be embedded</td>
</tr>
<tr>
<td><strong>Tool Misuse &amp; Privilege Escalation</strong></td>
<td>Agent uses granted tool access to perform unintended actions</td>
<td>Agents often receive broad tool permissions to function; a compromised agent can chain tools into damaging sequences</td>
</tr>
<tr>
<td><strong>Memory Poisoning</strong></td>
<td>Adversary corrupts the agent&#8217;s persistent memory or context</td>
<td>Long-running agents with persistent state can have their learned context manipulated over time</td>
</tr>
<tr>
<td><strong>Cascading Failures</strong></td>
<td>One compromised agent propagates bad decisions through a multi-agent system</td>
<td>Agentic architectures often chain agent outputs as inputs to other agents — one bad link breaks the chain</td>
</tr>
<tr>
<td><strong>Supply Chain Attacks</strong></td>
<td>Malicious code introduced through agent components, MCP servers, or skill libraries</td>
<td>The agent ecosystem relies heavily on community-built components with inconsistent vetting</td>
</tr>
</tbody>
</table>
<p>The <a href="https://arxiv.org/pdf/2603.09002" target="_blank" rel="noopener nofollow">arXiv paper &#8220;Security Considerations for Multi-agent Systems&#8221;</a> (March 2026) formalized many of these vectors. And real-world incidents are already validating the theory: the Barracuda Security report identified 43 different agent framework components with embedded vulnerabilities introduced via supply chain compromise.</p>
<p>The CyberAgents Exchange AI Inspector directly addresses the supply chain attack vector — the one that is hardest for individual teams to defend against, because it requires vetting components you did not write and may not fully understand.</p>
<h2 id="openai-daybreak-context">The OpenAI Daybreak Defense Network Context</h2>
<p>The Tenable-OpenAI collaboration did not happen in isolation. It sits inside a much larger strategic move by OpenAI to position frontier AI as a defensive cybersecurity asset.</p>
<p>On September 3 — the same day the AI Inspector was announced — <a href="https://openai.com/index/expanding-daybreak-as-the-cyber-defense-window-narrows/" target="_blank" rel="noopener nofollow">OpenAI expanded its Daybreak program</a> with a $1 billion global commitment for subsidized model access, training, technical support, and partnerships focused on defensive cybersecurity. The Daybreak Defense Network now includes more than 35 enterprise products and partner-operated services from companies including Accenture, IBM, Palo Alto Networks, CrowdStrike, Cisco, Sophos, Fortinet, and Cloudflare.</p>
<p>At the same summit, OpenAI announced that its upcoming GPT-6 Astra model had reached &#8220;Critical&#8221; cybersecurity capabilities under the company&#8217;s Preparedness Framework, meaning it can perform sophisticated cyber operations requiring new safeguards. OpenAI also committed $10 million in API credits for defensive cybersecurity grants.</p>
<p>Tenable&#8217;s participation in the Daybreak Defense Network is what gave it access to the specialized GPT cyber models powering Layer 1 of the AI Inspector. This matters architecturally: the Inspector&#8217;s automated analysis is not running a general-purpose LLM — it is using models specifically fine-tuned for cybersecurity reasoning, vulnerability discovery, and code analysis.</p>
<p>For Agentic AI Architects, the broader signal is equally important: the major AI labs are now investing heavily in the defensive side of the security equation, not just building more powerful models. OpenAI, Anthropic (which supported Tenable&#8217;s SWARM event at Black Hat), and Google all announced cybersecurity-specific model variants and programs in the same week. The infrastructure for securing agent ecosystems is finally being built at the same pace as the agent ecosystems themselves.</p>
<h2 id="component-types">What Gets Inspected: Agents, Skills, MCP Servers, and Playbooks</h2>
<p>Understanding what each component type is — and why each presents different security challenges — is essential for using the AI Inspector effectively.</p>
<h3>AI Agents</h3>
<p>In the CyberAgents Exchange context, an agent is an autonomous AI tool that performs security tasks end-to-end: scanning for vulnerabilities, triaging alerts, writing remediation scripts, or orchestrating incident response. Agents typically have tool access (the ability to call APIs, read files, execute commands) and may have persistent memory.</p>
<p><strong>Security concern:</strong> An agent&#8217;s tool access is its attack surface. A compromised or poorly designed agent can use legitimate tools for malicious purposes — the same API that reads vulnerability data can exfiltrate it.</p>
<h3>Skills</h3>
<p>Skills are reusable capabilities — smaller, composable units of functionality that extend an AI coding assistant or agent. Think of them as plugins: a CVE lookup skill, a log parser skill, a compliance checker skill. They are designed to be composed into larger workflows.</p>
<p><strong>Security concern:</strong> Skills often run with the permissions of the agent that invokes them. A malicious skill imported into a trusted agent inherits that agent&#8217;s access — a classic confused deputy attack. If you&#8217;ve worked with <a href="https://rpabotsworld.com/what-is-mcp-server-ai-agents/">MCP servers in agentic AI systems</a>, you&#8217;ve seen how tool-level permissions can cascade.</p>
<h3>MCP Servers</h3>
<p>Model Context Protocol servers connect AI systems with external resources and applications — databases, SIEMs, ticketing systems, threat feeds. MCP has become the de facto standard for <a href="https://rpabotsworld.com/a2a-joins-mcp-aaif-agentic-ai-architects-guide/">agent-to-tool connectivity</a>, and the security of MCP servers is now a critical concern as more enterprise systems expose MCP endpoints.</p>
<p><strong>Security concern:</strong> An MCP server is a bridge between an AI agent and a production system. A compromised MCP server can feed poisoned data to agents (triggering incorrect automated responses), exfiltrate data from production systems, or serve as a lateral movement vector into systems that were previously air-gapped from AI workloads.</p>
<h3>Multi-Agent Playbooks</h3>
<p>Playbooks are structured workflows where several agents perform separate roles to complete a larger task — for example, an incident response playbook where one agent triages, another enriches with threat intelligence, a third drafts a containment script, and a fourth manages stakeholder communication.</p>
<p><strong>Security concern:</strong> Playbooks introduce emergent risks that do not exist in any individual component. The output of Agent A becomes the input to Agent B, creating data flow paths that may not have been anticipated during design. If you&#8217;ve worked with <a href="https://rpabotsworld.com/why-agentic-automation-fails/">enterprise agentic automation programs</a>, you&#8217;ve seen how these cascading dependencies can amplify a single point of failure.</p>
<h2 id="for-architects">What This Means for Agentic AI Architects</h2>
<p>The CyberAgents Exchange AI Inspector is not just a security team&#8217;s tool — it is an architecture decision. Here is how to think about it at each layer of your agent system design.</p>
<h3>Component Selection: Trust but Verify</h3>
<p>If you are building enterprise agent systems that incorporate open-source components — and in 2026, virtually everyone is — the AI Inspector gives you a structured way to vet those components before they enter your architecture. This is analogous to how mature software teams use Snyk or Dependabot for dependency scanning, but applied to the specific risk profile of AI agent components.</p>
<p><strong>Practical step:</strong> Before integrating any community-built MCP server or agent skill into your system, check whether it is listed on the CyberAgents Exchange and whether it has passed the AI Inspector review. If it is not listed, consider submitting it yourself — the review process is free, and the three-layer inspection will surface risks you might miss in a manual code review.</p>
<h3>Architecture Patterns: Defense in Depth for Agents</h3>
<p>The AI Inspector addresses one layer of agent security — component vetting before deployment. But a production-grade agent architecture needs defense at every layer. Integrate the Inspector into a broader security posture:</p>
<ul>
<li><strong>Pre-deployment:</strong> CyberAgents Exchange AI Inspector (component vetting)</li>
<li><strong>Runtime:</strong> Least-privilege tool access, input validation on all agent-to-agent data flows, output guardrails</li>
<li><strong>Monitoring:</strong> Agent observability (LangSmith, Langfuse, or vendor-specific tools) to detect anomalous tool usage or data access patterns</li>
<li><strong>Governance:</strong> Enterprise agent governance frameworks — see coverage of <a href="https://rpabotsworld.com/servicenow-ai-control-tower-enterprise-agent-governance-guide/">ServiceNow AI Control Tower</a> and <a href="https://rpabotsworld.com/aws-dogwood-temporal-policies-agentcore-agent-governance-guide/">AWS Dogwood &amp; AgentCore</a> for platform-level governance approaches</li>
</ul>
<h3>Supply Chain Security Is Now Your Problem</h3>
<p>The shift from monolithic RPA bots to composable agent architectures means the supply chain attack surface has exploded. A UiPath bot is a single artifact you deploy from a controlled package. A modern agent system might incorporate three MCP servers, five skills, two sub-agents, and a playbook — each from a different source, each with its own dependency tree.</p>
<p>The AI Inspector is the first tool purpose-built for this new supply chain reality. But it only works if architects treat agent component vetting as a first-class architecture concern, not a security team afterthought. The <a href="https://rpabotsworld.com/rpa-to-agentic-ai-transition-guide/">transition from RPA to agentic AI</a> requires rethinking security as fundamentally as it requires rethinking process design.</p>
<h2 id="getting-started">Getting Started: Contributing and Consuming</h2>
<h3>Using the Exchange</h3>
<p>The CyberAgents Exchange is live at <a href="https://exchange.tenable.com/" target="_blank" rel="noopener nofollow">exchange.tenable.com</a>. You can browse all 100+ components by type (agents, skills, MCP servers, playbooks), search by keyword or tag, and view each component&#8217;s metadata including author, license, and community status. No Tenable account or product license is required to browse or consume.</p>
<h3>Contributing a Component</h3>
<p>To list your own agent, skill, MCP server, or playbook on the Exchange:</p>
<ol>
<li>Fork the <a href="https://github.com/tenable/cyberagents-exchange" target="_blank" rel="noopener nofollow">tenable/cyberagents-exchange</a> repository on GitHub</li>
<li>Create a Markdown file in the appropriate directory (<code>agents/</code>, <code>skills/</code>, <code>mcp-servers/</code>, or <code>playbooks/</code>) following the template format</li>
<li>Fill in the YAML frontmatter with your component&#8217;s metadata and write a description in the Markdown body</li>
<li>Run <code>validator.py</code> to check schema compliance</li>
<li>Submit a pull request — the Tenable team reviews submissions for quality and completeness</li>
</ol>
<p>Once listed, your component becomes eligible for the AI Inspector review process when it becomes generally available in September 2026.</p>
<h3>Integrating Into Your Development Workflow</h3>
<p>For teams building agent systems at scale, the recommended approach is to make the CyberAgents Exchange a gate in your component adoption process:</p>
<ol>
<li><strong>Discovery:</strong> Team member identifies a useful open-source agent component</li>
<li><strong>Registry check:</strong> Verify whether the component is listed on the CyberAgents Exchange</li>
<li><strong>Inspection review:</strong> If listed, check the AI Inspector status; if not listed, submit it for review</li>
<li><strong>Internal review:</strong> Complement the Exchange review with your own threat model specific to your environment</li>
<li><strong>Deployment:</strong> Deploy with least-privilege configuration and runtime monitoring</li>
</ol>
<h2 id="comparison">How CyberAgents Exchange Compares to Other Agent Registries</h2>
<p>The CyberAgents Exchange is not the only agent component registry, but it is the only one with a security-first design. Here is how it compares:</p>
<table>
<thead>
<tr>
<th>Feature</th>
<th>CyberAgents Exchange</th>
<th>Hugging Face Spaces</th>
<th>LangChain Hub</th>
<th>MCP Server Registries</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Focus</strong></td>
<td>Cybersecurity agents and tools</td>
<td>General AI models and demos</td>
<td>LangChain prompts and chains</td>
<td>MCP server listings</td>
</tr>
<tr>
<td><strong>Security review</strong></td>
<td>Three-layer AI Inspector (AI + platform + human)</td>
<td>Community flagging only</td>
<td>No formal review</td>
<td>Varies by registry</td>
</tr>
<tr>
<td><strong>Open source</strong></td>
<td>Yes (GitHub-hosted directory)</td>
<td>Mixed</td>
<td>Mixed</td>
<td>Varies</td>
</tr>
<tr>
<td><strong>Component types</strong></td>
<td>Agents, skills, MCP servers, playbooks</td>
<td>Models, datasets, spaces</td>
<td>Prompts, chains, tools</td>
<td>MCP servers only</td>
</tr>
<tr>
<td><strong>Enterprise backing</strong></td>
<td>Tenable (NASDAQ: TENB) + OpenAI + SentinelOne + Recorded Future</td>
<td>Hugging Face</td>
<td>LangChain Inc.</td>
<td>Community-led</td>
</tr>
<tr>
<td><strong>Code transparency</strong></td>
<td>Full (links to contributor repos)</td>
<td>Full</td>
<td>Full</td>
<td>Varies</td>
</tr>
<tr>
<td><strong>Cost</strong></td>
<td>Free</td>
<td>Free tier available</td>
<td>Free</td>
<td>Free</td>
</tr>
</tbody>
</table>
<p>The key differentiator is the AI Inspector. Other registries index components; the CyberAgents Exchange inspects them. For enterprise architects who need to justify component selection to security teams and compliance reviewers, this distinction matters.</p>
<h2 id="faqs">Frequently Asked Questions</h2>
<h3>Do I need a Tenable product license to use the CyberAgents Exchange?</h3>
<p>No. The CyberAgents Exchange is free, open-source, and vendor-agnostic. You can browse, consume, and contribute components without any Tenable license. The AI Inspector uses Tenable One AI Exposure in its second layer, but the review results are available to all Exchange users.</p>
<h3>Can I submit non-cybersecurity AI agents to the Exchange?</h3>
<p>The Exchange is specifically designed for cybersecurity-focused components. General-purpose agents, skills, or MCP servers that do not have a security function would not be accepted. For general agent registries, consider Hugging Face Spaces or LangChain Hub.</p>
<h3>Is the AI Inspector available now?</h3>
<p>As of the September 3, 2026 announcement, the AI Inspector is expected to be generally available in September 2026. The CyberAgents Exchange itself has been live since August 6, 2026 and is accepting contributions now.</p>
<h3>How does the AI Inspector handle false positives?</h3>
<p>The three-layer design is specifically intended to minimize false positives. Layer 1 (AI scan) casts a wide net; Layer 2 (Tenable One platform) applies production-context analysis; Layer 3 (human researcher review) applies judgment. Components flagged by the AI layer but cleared by human review are listed with their full review history for transparency.</p>
<h3>Does the AI Inspector review my proprietary agent code?</h3>
<p>The Exchange is a directory — it indexes component metadata and links to the contributor&#8217;s own repository. The AI Inspector reviews code that is publicly available on GitHub. If your agent is proprietary and not hosted in a public repository, it would not be submitted to the Exchange. For proprietary agent security review, Tenable One AI Exposure provides a separate, private analysis path.</p>
<h2 id="key-takeaways">Key Takeaways</h2>
<ul>
<li><strong>The CyberAgents Exchange AI Inspector</strong> is the first three-layer security review pipeline for AI agent components, combining OpenAI GPT cyber models, Tenable One AI Exposure, and human expert review.</li>
<li><strong>It addresses the supply chain attack vector</strong> — the hardest to defend against individually — by creating a structured, repeatable vetting process for community-built agents, skills, MCP servers, and multi-agent playbooks.</li>
<li><strong>The CyberAgents Exchange is free and open-source</strong>, with 100+ community-submitted components and backing from Tenable, OpenAI, SentinelOne, and Recorded Future.</li>
<li><strong>Enterprise agent adoption is outrunning security readiness:</strong> 62% of organizations experiment with agents but only 10% have a mature non-human identity strategy. The Inspector helps close this gap.</li>
<li><strong>For architects, component vetting is now a first-class design concern</strong> — not a security team afterthought. Integrate the Exchange as a gate in your agent component adoption workflow.</li>
<li><strong>The broader context matters:</strong> OpenAI&#8217;s $1 billion Daybreak expansion, Anthropic&#8217;s SWARM support, and Google&#8217;s cyber model programs signal that AI labs are finally investing in defensive infrastructure at scale.</li>
</ul>
<h2 id="references">References</h2>
<ol>
<li>Tenable. &#8220;Tenable Uses OpenAI GPT Cyber Models to Help Defenders Inspect Community-Built AI Components.&#8221; Press release, September 3, 2026. <a href="https://www.tenable.com/press-releases/tenable-uses-openai-gpt-cyber-models-to-help-defenders-inspect-community-built-ai-components" target="_blank" rel="noopener nofollow">tenable.com</a></li>
<li>Tenable. &#8220;Tenable Launches Industry&#8217;s First Open-Source AI Agent Exchange.&#8221; Press release, August 6, 2026. <a href="https://www.tenable.com/press-releases/tenable-launches-industrys-first-open-source-ai-agent-exchange" target="_blank" rel="noopener nofollow">tenable.com</a></li>
<li>McKinsey &amp; Company. &#8220;The State of AI: Global Survey 2026.&#8221; <a href="https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai" target="_blank" rel="noopener nofollow">mckinsey.com</a></li>
<li>OpenAI. &#8220;Expanding Daybreak as the Cyber Defense Window Narrows.&#8221; September 3, 2026. <a href="https://openai.com/index/expanding-daybreak-as-the-cyber-defense-window-narrows/" target="_blank" rel="noopener nofollow">openai.com</a></li>
<li>CNBC. &#8220;OpenAI begins rolling out Astra model after warning of its advanced cyber capabilities.&#8221; September 3, 2026. <a href="https://www.cnbc.com/2026/09/03/open-ai-astra-gpt-6-cyber.html" target="_blank" rel="noopener nofollow">cnbc.com</a></li>
<li>Aembit. &#8220;6 Agentic AI Security Risks to Monitor in 2026.&#8221; <a href="https://aembit.io/blog/agentic-ai-cybersecurity-risks-security-guide/" target="_blank" rel="noopener nofollow">aembit.io</a></li>
<li>Salt Security. &#8220;Top Security Risks of AI Agents in 2026.&#8221; <a href="https://salt.security/blog/top-security-risks-of-ai-agents" target="_blank" rel="noopener nofollow">salt.security</a></li>
<li>arXiv. &#8220;Security Considerations for Multi-agent Systems.&#8221; March 2026. <a href="https://arxiv.org/pdf/2603.09002" target="_blank" rel="noopener nofollow">arxiv.org</a></li>
<li>GitHub. &#8220;tenable/cyberagents-exchange.&#8221; <a href="https://github.com/tenable/cyberagents-exchange" target="_blank" rel="noopener nofollow">github.com</a></li>
<li>IT Brief UK. &#8220;Tenable &amp; OpenAI launch AI inspector for cyber agents.&#8221; <a href="https://itbrief.co.uk/story/tenable-openai-launch-ai-inspector-for-cyber-agents" target="_blank" rel="noopener nofollow">itbrief.co.uk</a></li>
</ol>
]]></content:encoded>
					
					<wfw:commentRss>https://rpabotsworld.com/tenable-cyberagents-exchange-ai-inspector-agent-security-guide/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>Atlassian Rovo Agents, MCP Server, and AMP: The Complete Guide for Agentic AI Architects (2026)</title>
		<link>https://rpabotsworld.com/atlassian-rovo-mcp-agentic-multiplayer-protocol-guide/</link>
					<comments>https://rpabotsworld.com/atlassian-rovo-mcp-agentic-multiplayer-protocol-guide/#respond</comments>
		
		<dc:creator><![CDATA[Satish Prasad]]></dc:creator>
		<pubDate>Mon, 26 Oct 2026 05:09:26 +0000</pubDate>
				<category><![CDATA[AI Agents & Frameworks]]></category>
		<category><![CDATA[MCP]]></category>
		<guid isPermaLink="false">https://rpabotsworld.com/?p=32431</guid>

					<description><![CDATA[Atlassian Rovo agents now ship with 220+ MCP tools, real-time agent presence (AMP), and multi-model routing.]]></description>
										<content:encoded><![CDATA[<h1>Atlassian Rovo Agents, MCP Server, and AMP: The Complete Guide for Agentic AI Architects (2026)</h1>
<p>A Jira ticket sits in &#8220;In Progress&#8221; for three days. The developer assigned to it has been in back-to-back meetings, the Confluence design spec was updated twice since they last read it, and the Bitbucket branch they started from is now four commits behind main. Nobody notices until standup.</p>
<p>Now imagine the same ticket, but an AI agent is watching. It reads the full comment thread, checks the latest Confluence spec, notices the branch drift, flags the blocker in a comment, and — if your admin has enabled it — drafts a fix and opens a pull request. The developer reviews the PR after their meetings, not the raw problem.</p>
<p>That is the shift Atlassian announced at <strong>Team &#8217;26 Europe</strong> in October 2026: agents are no longer assistants you visit in a chat window. They are <strong>teammates with identity, presence, and permissions</strong> that show up alongside humans in Jira boards, Confluence pages, and Bitbucket repositories. The technical machinery behind this — a rebuilt MCP server with 220+ tools processing over 15 million calls per day, a new Agentic Multiplayer Protocol (AMP) for human-agent collaboration, and an expanded OpenAI partnership layering GPT-6 Astra into a multi-model routing architecture — represents one of the most complete enterprise agent platforms shipping today.</p>
<p>For agentic AI architects and RPA professionals, this matters because Atlassian is not building another chatbot. It is turning the tool where 300,000+ organizations already track their work into a <strong>first-class agent orchestration surface</strong> — and it is doing it with an open MCP architecture that connects to Claude, ChatGPT, Gemini, GitHub Copilot, Cursor, and any MCP-compatible client.</p>
<p>This guide breaks down every layer of the new stack: what shipped, what is in beta, what it costs, and how it fits into the broader agentic AI landscape that RPA teams are navigating right now.</p>
<h2>What Atlassian Announced at Team &#8217;26 Europe: The Full Picture</h2>
<p>The Team &#8217;26 Europe conference (October 6–7, 2026) landed five interconnected announcements that, taken together, redefine how agents operate inside enterprise work management:</p>
<table border="1" cellpadding="8" cellspacing="0" style="width:100%; border-collapse:collapse;">
<thead>
<tr style="background-color:#f5f5f5;">
<th>Announcement</th>
<th>What It Is</th>
<th>Status</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Rebuilt MCP Server (v2)</strong></td>
<td>220+ tools across Jira, Confluence, Bitbucket, Loom, Goals — up to 25% fewer tokens</td>
<td>Generally Available</td>
</tr>
<tr>
<td><strong>Agentic Multiplayer Protocol (AMP)</strong></td>
<td>Framework for agents to appear alongside humans in real-time presence bars and cursors</td>
<td>Rolling out</td>
</tr>
<tr>
<td><strong>Rovo Work</strong></td>
<td>Multi-step agent tasks that can run for hours in an admin-governed sandbox</td>
<td>Early Access</td>
</tr>
<tr>
<td><strong>OpenAI Partnership Expansion</strong></td>
<td>GPT-6 Astra and GPT-5.6 series added to Rovo&#8217;s multi-model router</td>
<td>Available (admin opt-in)</td>
</tr>
<tr>
<td><strong>Agent Accounts &amp; Identity</strong></td>
<td>Scoped agent identities that persist like user accounts, with full audit trails</td>
<td>Preview / Coming Soon</td>
</tr>
</tbody>
</table>
<p>Each of these layers reinforces the others. The MCP server provides the tool surface. AMP provides the collaboration model. Rovo Work provides the execution runtime. The OpenAI partnership (alongside existing Anthropic Claude and Google Gemini integrations) provides the reasoning layer. And agent accounts provide the identity and governance layer that enterprise buyers demand.</p>
<p>Let us examine each one.</p>
<h2>The Rebuilt Atlassian MCP Server: 220+ Tools and 15 Million Daily Calls</h2>
<p>The MCP server is the backbone of Atlassian&#8217;s agentic strategy. Rebuilt from the ground up since the original December 2025 ChatGPT integration, it is now the single access point through which any AI agent — first-party or third-party — reads, writes, and acts on Atlassian data.</p>
<h3>Architecture and Endpoint</h3>
<p>The server is <strong>cloud-hosted by Atlassian</strong> at <code>https://mcp.atlassian.com/v2/mcp</code> (the legacy <code>/v1/mcp</code> and <code>/sse</code> endpoints remain supported, with v1 tools migrating to v2 automatically on March 1, 2027). Authentication uses <strong>OAuth 2.1 with PKCE</strong> by default, with API token authentication available for headless and CI/CD use cases — though an org admin must explicitly enable API token auth first.</p>
<p>A key architectural decision is <strong>progressive tool discovery</strong>. Rather than dumping all 220+ tool definitions into a client&#8217;s context window on connection, the server advertises a small set of primary tools and exposes the rest through a dedicated <code>discover</code> tool. This conserves context window budget (critical for cost-conscious agentic deployments) and allows Atlassian to add new tools without requiring clients to reconnect.</p>
<p>Anthropic&#8217;s head of enterprise product, Scott White, called it <em>&#8220;one of the most-used enterprise MCP integrations on Claude&#8221;</em> — and the 25% token reduction claim was benchmarked specifically on Claude models.</p>
<h3>Tool Coverage by Product</h3>
<p>The rebuilt server spans Atlassian&#8217;s full product suite:</p>
<table border="1" cellpadding="8" cellspacing="0" style="width:100%; border-collapse:collapse;">
<thead>
<tr style="background-color:#f5f5f5;">
<th>Product</th>
<th>Capabilities</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Jira</strong></td>
<td>Search, create, edit, transition, and comment on work items. Query boards, sprints, versions, and metadata.</td>
</tr>
<tr>
<td><strong>Confluence</strong></td>
<td>Read, create, and update pages, live databases, and whiteboards. CQL search. Manage comments, attachments, permissions, and spaces.</td>
</tr>
<tr>
<td><strong>Bitbucket Cloud</strong></td>
<td>Repositories, pull requests, pipelines, deployments, environments, branch state. Semantic code exploration via Code Search.</td>
</tr>
<tr>
<td><strong>Loom</strong></td>
<td>Retrieve recordings, transcripts, comments, AI briefs, meeting action items, and folders.</td>
</tr>
<tr>
<td><strong>Jira Service Management</strong></td>
<td>Query and update ops alerts, on-call schedules, and team information.</td>
</tr>
<tr>
<td><strong>Goals, Projects, Focus Areas</strong></td>
<td>Search, read, create, and update strategy entities.</td>
</tr>
<tr>
<td><strong>Teamwork Graph &amp; Rovo Search</strong></td>
<td>Cross-app context retrieval, semantic search across all connected products and third-party tools.</td>
</tr>
</table>
<p>The <strong>Teamwork Graph</strong> deserves special attention. Described by Atlassian as a layer mapping more than 150 billion objects and relationships across Jira, Confluence, Bitbucket, Loom, and 75+ third-party tools, it is what gives agents pre-linked context on the first turn. When an agent picks up a Jira ticket, the Teamwork Graph can surface the related Confluence spec, the Bitbucket branch, the Loom walkthrough, and the team members who last touched each artifact — without the agent having to make separate discovery calls.</p>
<h3>Security and Admin Controls</h3>
<p>Enterprise governance is where Atlassian&#8217;s MCP server differentiates from most open-source MCP implementations:</p>
<ul>
<li><strong>Permission inheritance:</strong> Every tool call runs under the authenticated user&#8217;s existing Atlassian permissions. Agents never get more access than the user already holds.</li>
<li><strong>Risk-tiered actions:</strong> Read, write, search, and destructive actions are separated into distinct permission groups. Clients can require human confirmation for higher-risk steps.</li>
<li><strong>Atlassian Guard integration:</strong> Data Security Policies and Data Loss Prevention controls apply to MCP connections on Guard Standard and above.</li>
<li><strong>Admin Console controls:</strong> Domain allowlists, IP allowlists, per-scope read/write/search permission toggles, and organization-wide audit logging.</li>
<li><strong>Full audit trail:</strong> Every agent action is logged — on all plan tiers, not just Enterprise.</li>
</ul>
<p>One nuance worth noting for security teams: the domain allowlist and MCP data-security controls apply only to OAuth connections. API-token and enterprise managed authentication (currently in beta) bypass them. This is a known gap that architects should factor into their security posture.</p>
<h3>Supported Clients</h3>
<p>The MCP server works with any MCP-compatible client, including: OpenAI ChatGPT, Claude (claude.ai, Desktop, and Code), Cursor, VS Code (GitHub Copilot), GitHub Copilot CLI, Google Gemini, Amazon Quick Suite, Kiro, Docker, and Windsurf. One-click install is available for Cursor, VS Code, ChatGPT, and Claude. Local clients use the <code>mcp-remote</code> proxy (Node.js v18+ required).</p>
<p>The open-source server repository (<a href="https://github.com/atlassian/atlassian-mcp-server" target="_blank" rel="noopener nofollow">atlassian/atlassian-mcp-server</a>) carries an Apache-2.0 license and has accumulated over 1,000 GitHub stars.</p>
<h2>AMP: The Agentic Multiplayer Protocol</h2>
<p>If the MCP server is the <em>how</em> of agent access, AMP is the <em>where</em> of agent presence. Atlassian describes AMP as <em>&#8220;the collection of human-centric experiences, technologies, and patterns that shape how humans and agents work together.&#8221;</em></p>
<p>In practical terms, AMP means:</p>
<ul>
<li><strong>Agents appear in real-time presence bars and cursors</strong> alongside human teammates — on Jira boards, Confluence pages, and Rovo Chat.</li>
<li><strong>Registration once, deployment anywhere:</strong> You register an agent in Rovo Studio and then choose where it appears — a specific Jira board, a Confluence space, or Rovo Chat.</li>
<li><strong>Identity options:</strong> Agents can run as a user (inheriting that user&#8217;s permissions) or use dedicated service accounts with scoped permissions.</li>
</ul>
<p>This is a conceptual shift from how most enterprise agent platforms work today. In platforms like <a href="https://rpabotsworld.com/salesforce-agentforce-job-ready-agents-long-horizon-runtime-guide-3/">Salesforce Agentforce</a>, agents are invoked through dedicated agent surfaces. In Atlassian&#8217;s vision, agents are <em>co-present</em> — they show up in the same UI where humans are already working, with the same visual indicators of activity.</p>
<p>The implications for RPA teams are significant. Traditional RPA bots operate behind the scenes, processing queues and triggering workflows in the background. AMP positions agents as visible, accountable teammates whose actions are transparent to everyone on the board. This is closer to the &#8220;attended automation&#8221; model that RPA architects have long sought — but with the reasoning capabilities of frontier LLMs instead of rule-based scripts.</p>
<h2>Rovo Work: Multi-Step Agent Tasks That Run for Hours</h2>
<p>Rovo Work is a new Rovo Chat mode designed for <strong>long-running, multi-step tasks</strong> that can execute for hours in an admin-governed sandbox. While details remain limited (it was not visible on test sites at the time of Team &#8217;26 Europe), Atlassian positions it as the bridge between simple chat-based agent interactions and full autonomous workflow execution.</p>
<p>Key characteristics:</p>
<ul>
<li>Tasks run in a <strong>sandboxed environment</strong> with admin-defined boundaries</li>
<li>Falls under <strong>premium AI interactions</strong> with variable pricing based on compute consumed</li>
<li>Designed for complex, multi-tool workflows — triaging a backlog, auditing a Confluence space for outdated content, generating release notes from a sprint&#8217;s worth of commits</li>
</ul>
<p>For agentic AI architects building on Atlassian, Rovo Work represents the execution layer that sits between ad-hoc agent queries (which consume 10 Rovo credits per request) and fully autonomous agent workflows (still on the roadmap). It is where the comparison to UiPath&#8217;s long-running orchestrated processes or <a href="https://rpabotsworld.com/automation-anywhere-vs-uipath-agentic-procure-to-pay-2026/">Automation Anywhere&#8217;s agentic process automation</a> becomes most direct.</p>
<h2>The OpenAI Partnership: GPT-6 Astra Inside a Multi-Model Router</h2>
<p>The expanded OpenAI partnership adds <strong>GPT-6 Astra</strong> (released September 3, 2026) and the <strong>GPT-5.6 series</strong> to Rovo&#8217;s model palette. But the architecturally significant detail is that Atlassian is <em>not</em> making this an exclusive arrangement.</p>
<h3>Rovo&#8217;s Multi-Model Routing</h3>
<p>Rovo uses an <strong>internal AI gateway</strong> that dynamically routes tasks across providers based on three factors: capability (which model handles this task class best), speed (latency requirements), and cost (token economics). GPT-6 Astra is available but <em>not</em> the default model — Atlassian Intelligence already runs on Anthropic Claude, Google Gemini 3 Flash, and other providers.</p>
<p>The VentureBeat reporting on this partnership noted that Anthropic&#8217;s Claude Fable 5.1 scored higher than Astra on the Artificial Analysis Intelligence Index — a measure OpenAI itself publishes. Atlassian&#8217;s multi-model approach means it can route each request to whichever model offers the best quality-cost tradeoff for that specific task.</p>
<h3>GPT-6 Astra: Capabilities and Cautions</h3>
<p>GPT-6 Astra is priced at <strong>$10 per million input tokens</strong> and <strong>$50 per million output tokens</strong>. It is OpenAI&#8217;s first model to reach the &#8220;Critical&#8221; cybersecurity threshold under its Preparedness Framework, scoring 100% on one exploit benchmark and surfacing two previously unknown zero-day vulnerabilities during testing.</p>
<p>For enterprise deployments, important cautions apply:</p>
<ul>
<li>Safety checks can <em>&#8220;slow, pause, or stop legitimate work&#8221;</em> — including defensive security tasks</li>
<li>OpenAI acknowledges that Astra&#8217;s written reasoning is <em>&#8220;harder to monitor than its predecessor&#8217;s,&#8221;</em> which it considers <em>&#8220;a serious regression&#8221;</em></li>
<li>Astra currently refuses advanced offensive tasks (e.g., writing proof-of-concept exploits); OpenAI plans to loosen this for vetted defensive workflows</li>
<li>Enterprise access is <strong>off by default</strong> until an admin explicitly enables it</li>
<li>Zero Data Retention is available for &#8220;eligible&#8221; API customers, though what &#8220;eligible&#8221; means is not fully defined</li>
</ul>
<h2>Agent Accounts and Identity: The Governance Layer</h2>
<p>Perhaps the most architecturally significant announcement for enterprise RPA teams is the introduction of <strong>agent accounts</strong> — dedicated, persistent identities for AI agents that work like user accounts but with scoped permissions and full audit trails.</p>
<h3>How Agent Identity Works</h3>
<p>In Rovo Studio, administrators can configure agents with two identity models:</p>
<ul>
<li><strong>User&#8217;s account:</strong> The agent inherits the permissions of the human who invoked it. Actions are attributed to that user.</li>
<li><strong>Agent&#8217;s account:</strong> The agent operates under a dedicated service account with its own scoped permissions. Actions are attributed to the agent identity.</li>
</ul>
<p>Agents can be <strong>assigned Jira work items</strong>, <strong>@mentioned in comments</strong>, and <strong>move through Jira workflow states</strong> just like human teammates. They show up as assignees in sprint boards and their session status (stuck, waiting, or complete) is visible to the team.</p>
<p>This is a direct answer to one of the thorniest governance questions in enterprise agentic AI: <em>who did what?</em> When an agent modifies a Confluence page, transitions a Jira ticket, or opens a Bitbucket PR, the audit trail shows which agent acted, under which identity model, and with which permissions — at every plan tier.</p>
<h3>Cross-Vendor Identity: The Unsolved Problem</h3>
<p>Analyst coverage from Futurum Group flags an important limitation: when a workflow spans Rovo and an outside orchestrator (such as Gemini Enterprise or <a href="https://rpabotsworld.com/microsoft-agent-framework-1-0-foundry-agentic-ai-architect-guide/">Microsoft&#8217;s Agent Framework</a>), the identity model breaks down. Which system&#8217;s identity governs the cross-boundary action? This remains unresolved — and it is precisely the kind of interoperability challenge that the <a href="https://rpabotsworld.com/a2a-joins-mcp-aaif-agentic-ai-architects-guide/">Agentic AI Foundation (AAIF) and the A2A protocol</a> are designed to address.</p>
<h2>Agents in Action: Jira and Confluence Use Cases</h2>
<p>The practical impact shows most clearly in specific use cases that are now generally available or in open beta:</p>
<h3>Jira: Agents as Assignees</h3>
<ul>
<li><strong>Coding agent assignment (GA):</strong> Engineering teams can assign Jira work items to Claude Code, Cursor, and GitHub Copilot. The agent reads the ticket context, checks related artifacts via the Teamwork Graph, and produces a PR for human review.</li>
<li><strong>Built-in Jira Coding Agent (GA):</strong> Included in every paid Jira plan at no extra cost. Turns work items into pull requests ready for review.</li>
<li><strong>Automation-triggered agent routing:</strong> An automation rule builder lets admins route specific work types (bug fixes, remediation tickets) to agents automatically when they enter a designated status.</li>
<li><strong>DX AI cost management:</strong> A reporting layer maps token consumption and agent spend data to engineering output metrics — answering the question &#8220;what did our AI agents actually produce?&#8221;</li>
</ul>
<h3>Confluence: Agents as Collaborators</h3>
<ul>
<li><strong>@mention agents on pages (open beta):</strong> You can tag agents from Lovable, Replit, Databricks, Gamma, and others on a Confluence page. The agent reads the page context and acts across connected tools.</li>
<li><strong>Remix with Rovo (beta):</strong> Converts selected Confluence text into charts, timelines, and infographics.</li>
<li><strong>Confluence Slides (beta):</strong> Rovo generates and edits presentations grounded in team work, presentable within Confluence.</li>
<li><strong>Create with Rovo in Jira (beta):</strong> Turns Confluence docs, meeting summaries, and email threads into Jira work items.</li>
</ul>
<h3>Cross-Product: The Teamwork Graph in Action</h3>
<p>The most compelling use cases span multiple products:</p>
<ul>
<li><strong>Meeting-to-work pipeline:</strong> A Loom recording is transcribed, decisions are extracted, and the agent drafts Jira work items and Confluence spec pages — all without a human doing the manual conversion.</li>
<li><strong>Ticket catch-up:</strong> An agent reads the full comment thread and field-level changelog of a Jira ticket, synthesizes the history, and presents a summary with key decisions highlighted.</li>
<li><strong>Safe page editing:</strong> When an agent edits a Confluence page, it diffs document versions and changes only the requested section. If someone else edits the page mid-task, the agent <em>pauses to avoid overwrites</em> and previews changes before publishing.</li>
</ul>
<h2>Pricing: What Rovo Agents Actually Cost</h2>
<p>One of the strongest aspects of Atlassian&#8217;s agent strategy is the pricing model: <strong>Rovo is bundled into all paid Jira, Confluence, JSM, and Teamwork Collection cloud subscriptions</strong> at no separate per-seat fee.</p>
<h3>Credit Allowances</h3>
<table border="1" cellpadding="8" cellspacing="0" style="width:100%; border-collapse:collapse;">
<thead>
<tr style="background-color:#f5f5f5;">
<th>Plan</th>
<th>Credits/User/Month (Jira/Confluence/JSM)</th>
<th>Credits/User/Month (Teamwork Collection)</th>
</tr>
</thead>
<tbody>
<tr>
<td>Standard</td>
<td>25</td>
<td>250</td>
</tr>
<tr>
<td>Premium</td>
<td>70</td>
<td>700</td>
</tr>
<tr>
<td>Enterprise</td>
<td>150</td>
<td>1,500</td>
</tr>
</tbody>
</table>
<h3>Credit Consumption</h3>
<table border="1" cellpadding="8" cellspacing="0" style="width:100%; border-collapse:collapse;">
<thead>
<tr style="background-color:#f5f5f5;">
<th>Action</th>
<th>Credits</th>
</tr>
</thead>
<tbody>
<tr>
<td>Rovo Chat request</td>
<td>10</td>
</tr>
<tr>
<td>Rovo Agent request</td>
<td>10</td>
</tr>
<tr>
<td>Deep Research request</td>
<td>100</td>
</tr>
<tr>
<td>Teamwork Graph / Rovo Search (per MCP call)</td>
<td>Up to 10</td>
</tr>
<tr>
<td>Rovo Search, inline summaries, chart insights</td>
<td>Free</td>
</tr>
</tbody>
</table>
<p>Credits are pooled across the organization, reset monthly with no rollover. As of October 2026, <strong>Atlassian is not billing for overage</strong> — requests above the pool still execute. Atlassian has committed to providing at least 90 days&#8217; notice before overage becomes billable, with an explicit opt-in required. Extra-usage billing is tentatively scheduled to start <strong>December 3, 2026</strong>, at a list price of $0.01 per credit.</p>
<p>For comparison, Salesforce Agentforce charges $2 per conversation for its standard agent interactions. Atlassian&#8217;s credit-based model is substantially more granular — and currently free above the included allowance.</p>
<h3>JSM-Specific Pricing</h3>
<ul>
<li><strong>Rovo Customer Service:</strong> $1 per resolution (charged only when the AI fully resolves a ticket without human escalation)</li>
<li><strong>Virtual Service Agent:</strong> 1,000 assisted conversations/month included (12,000/year on annual billing), with $0.30 per conversation overage</li>
</ul>
<h2>The Ecosystem: Third-Party Agents and Extensibility</h2>
<p>Atlassian is positioning its platform as an <strong>open agent marketplace</strong>, not a walled garden:</p>
<ul>
<li><strong>Third-party coding agents on the Atlassian Marketplace:</strong> Devin, Factory, and Warp are already available. OpenAI Codex is &#8220;coming soon.&#8221;</li>
<li><strong>Confluence agent integrations (open beta):</strong> Lovable, Replit, Databricks, and Gamma agents can be @mentioned on pages.</li>
<li><strong>Forge extensibility:</strong> Developers can build custom MCP tools with the <code>rovo:mcp</code> module (Preview since August 14, 2026) and add MCP capabilities to Marketplace apps.</li>
<li><strong>Bidirectional MCP with Google:</strong> Rovo is reachable from within Gemini Enterprise, and Google Gemini 3 Flash powers select Rovo capabilities.</li>
<li><strong>Over 3,000 internal users:</strong> Atlassian itself uses Codex through ChatGPT Enterprise, with its DX platform measuring delivery impact.</li>
</ul>
<h2>What Is on the Roadmap (Not Yet Shipped)</h2>
<p>Several capabilities were announced as coming but are not yet available:</p>
<ul>
<li><strong>Autonomous agents that pick up Jira work items</strong> without human assignment, run tests, and support multi-agent orchestration with human checkpoints</li>
<li><strong>Assigning tasks to AI agents</strong> with progress tracking and decision recording in Jira</li>
<li><strong>Deeper Codex integration</strong> in planning and code review workflows</li>
<li><strong>Syncing local agent session history</strong> back to team boards</li>
<li><strong>Agent Sessions (closed EAP):</strong> Linking local and cloud agent runs to work items</li>
<li><strong>Interactive PR Reviews (closed EAP)</strong></li>
<li><strong>Service Collection AI agents in Slack, Teams, and Claude (closed EAP)</strong></li>
<li><strong>Local EU inference:</strong> Restricting LLM processing to EU-hosted models (rolling out)</li>
<li><strong>Non-human identity inventory:</strong> A centralized view of all agent identities across the org</li>
</ul>
<h2>What This Means for RPA Teams</h2>
<p>For the audience that reads RPABOTS.WORLD — RPA developers, automation architects, and teams evaluating the move from traditional automation to agentic AI — Atlassian&#8217;s announcements carry several implications:</p>
<h3>1. The Work Management Layer Becomes the Agent Orchestration Layer</h3>
<p>Traditional RPA orchestrators (UiPath Orchestrator, Automation Anywhere Control Room) manage bot queues, schedules, and credentials. Atlassian is building a parallel orchestration layer — but the &#8220;queue&#8221; is your existing Jira backlog, the &#8220;schedule&#8221; is your sprint cadence, and the &#8220;credential&#8221; is an OAuth-scoped agent account. For organizations already on Atlassian, this is an agent runtime they do not need to buy separately.</p>
<h3>2. MCP Replaces Point-to-Point Integrations</h3>
<p>RPA teams are accustomed to building and maintaining connectors between their automation platform and enterprise tools. Atlassian&#8217;s MCP server provides a <strong>single, standardized integration point</strong> with 220+ pre-built tools, governed by existing user permissions. Instead of building a UiPath activity to read a Confluence page or an Automation Anywhere action to transition a Jira ticket, an agentic workflow can call the MCP server directly — from any MCP-compatible client.</p>
<h3>3. Agent Visibility Changes the Trust Equation</h3>
<p>One of the persistent challenges in RPA adoption is the &#8220;black box&#8221; problem: business users do not see what bots are doing until something goes wrong. AMP&#8217;s agent presence model — where agents appear in real-time alongside humans on Jira boards — addresses this directly. When a team can see that an agent is working on a ticket, has been stuck for 20 minutes, or has completed its task, the trust dynamic shifts from &#8220;I hope the bot ran&#8221; to &#8220;I can see what the agent is doing.&#8221;</p>
<h3>4. Credit-Based Pricing Aligns with RPA Consumption Models</h3>
<p>RPA pricing has moved from per-bot licensing to consumption-based models (UiPath&#8217;s per-transaction pricing, Automation Anywhere&#8217;s FTE-based model). Atlassian&#8217;s credit-based system follows the same trajectory — and with the current zero-overage policy, organizations can experiment extensively before committing to usage-based spend.</p>
<h2>Architect&#8217;s Checklist: Evaluating Atlassian&#8217;s Agent Platform</h2>
<p>Before adopting Rovo agents in production, agentic AI architects should evaluate these dimensions:</p>
<table border="1" cellpadding="8" cellspacing="0" style="width:100%; border-collapse:collapse;">
<thead>
<tr style="background-color:#f5f5f5;">
<th>Dimension</th>
<th>Current State</th>
<th>Watch For</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Tool Coverage</strong></td>
<td>220+ tools, strong on Jira/Confluence/Bitbucket</td>
<td>JSM support still &#8220;coming soon&#8221;; third-party tool coverage via Teamwork Graph</td>
</tr>
<tr>
<td><strong>Security</strong></td>
<td>OAuth 2.1, permission inheritance, Guard integration, audit logs</td>
<td>API token auth bypasses domain allowlists; prompt-injection risk acknowledged but not fully mitigated</td>
</tr>
<tr>
<td><strong>Model Flexibility</strong></td>
<td>Multi-model routing (OpenAI, Anthropic, Google)</td>
<td>No self-hosted model option; EU inference &#8220;rolling out&#8221; but not GA</td>
</tr>
<tr>
<td><strong>Cost Control</strong></td>
<td>Credits included in existing plans; overage not yet billed</td>
<td>December 2026 overage billing start; Rovo Work pricing unclear</td>
</tr>
<tr>
<td><strong>Extensibility</strong></td>
<td>Forge custom MCP tools, Marketplace apps</td>
<td>Forge <code>rovo:mcp</code> module still in Preview</td>
</tr>
<tr>
<td><strong>Observability</strong></td>
<td>Agent session visibility, DX cost management</td>
<td>Described by analysts as &#8220;shallow, closer to dashboards than production telemetry&#8221;</td>
</tr>
<tr>
<td><strong>Identity</strong></td>
<td>User accounts and agent accounts in Rovo Studio</td>
<td>Cross-vendor identity unresolved; non-human identity inventory not yet shipped</td>
</tr>
</tbody>
</table>
<h2>How Atlassian&#8217;s Agent Platform Compares</h2>
<p>Context matters. Here is how Atlassian&#8217;s approach compares to the other major enterprise agent platforms:</p>
<table border="1" cellpadding="8" cellspacing="0" style="width:100%; border-collapse:collapse;">
<thead>
<tr style="background-color:#f5f5f5;">
<th>Platform</th>
<th>Agent Surface</th>
<th>MCP Support</th>
<th>Model Strategy</th>
<th>RPA Integration</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Atlassian Rovo</strong></td>
<td>Jira, Confluence, Bitbucket (co-present)</td>
<td>220+ tools, open server</td>
<td>Multi-model router</td>
<td>No native RPA; complements existing RPA</td>
</tr>
<tr>
<td><strong>Salesforce Agentforce</strong></td>
<td>CRM-native agent surface</td>
<td>MCP support via partner integrations</td>
<td>Multi-model (Einstein)</td>
<td>MuleSoft-based automation</td>
</tr>
<tr>
<td><strong>Microsoft Copilot Studio</strong></td>
<td>M365, Teams, Power Platform</td>
<td>MCP via Agent Framework</td>
<td>Azure OpenAI + partners</td>
<td>Power Automate native</td>
</tr>
<tr>
<td><strong>Google Vertex AI Agents</strong></td>
<td>Google Workspace, Cloud Console</td>
<td>A2A protocol (Google-led)</td>
<td>Gemini family</td>
<td>Apigee + partner connectors</td>
</tr>
<tr>
<td><strong>ServiceNow</strong></td>
<td>Now Platform workflows</td>
<td>AI Orchestrator + partners</td>
<td>Multi-model</td>
<td>Flow Designer automation</td>
</tr>
</tbody>
</table>
<p>Atlassian&#8217;s differentiator is the combination of <strong>open MCP architecture</strong> (any client, any model), <strong>co-present agents</strong> (not a separate agent surface), and <strong>bundled pricing</strong> (no separate agent SKU). Its weakness is the lack of native RPA or workflow automation capabilities — it depends on partners and the Forge platform for anything beyond Jira/Confluence/Bitbucket actions.</p>
<h2>Getting Started: Connecting Your First Agent</h2>
<p>For teams ready to experiment, the quickest path is:</p>
<ol>
<li><strong>Verify prerequisites:</strong> You need an Atlassian Cloud site on a paid plan (Standard, Premium, or Enterprise). An admin must verify the business domain — generic email domains (gmail.com, etc.) cannot enable Rovo.</li>
<li><strong>Enable Atlassian Intelligence:</strong> An org admin enables AI features in the Admin Console.</li>
<li><strong>Connect an MCP client:</strong> Use the one-click install for Claude Desktop, Cursor, VS Code, or ChatGPT. For other clients, configure with the endpoint <code>https://mcp.atlassian.com/v2/mcp</code> and complete the OAuth 2.1 flow.</li>
<li><strong>Start with read-only:</strong> Begin with search and read operations (ticket lookups, Confluence searches, code exploration) before enabling write operations.</li>
<li><strong>Monitor credits:</strong> Watch credit consumption in the Admin Console. Notifications trigger at 80% and 100% of your monthly allowance.</li>
</ol>
<h2>Frequently Asked Questions</h2>
<h3>Is Rovo a separate product I need to buy?</h3>
<p>No. Rovo is bundled into all paid Jira, Confluence, JSM, and Teamwork Collection cloud subscriptions. There is no separate per-seat fee. You get a monthly credit allowance based on your plan tier (25–1,500 credits per user per month depending on plan and bundle).</p>
<h3>Can I use Rovo agents with Claude, not just ChatGPT?</h3>
<p>Yes. The Atlassian MCP server works with any MCP-compatible client — Claude Desktop, Claude Code, ChatGPT, Gemini, Cursor, VS Code, GitHub Copilot, and others. Atlassian&#8217;s Rovo router itself uses Claude models alongside OpenAI and Google models internally.</p>
<h3>Does the MCP server replace the need for Jira/Confluence REST APIs in my automations?</h3>
<p>For agentic workflows, largely yes — the MCP server provides a higher-level interface with 220+ pre-built tools that include permission inheritance and audit logging. For traditional API integrations (webhooks, scheduled syncs, custom app development), the REST APIs remain the right choice.</p>
<h3>What happens when an agent and a human edit the same Confluence page simultaneously?</h3>
<p>Atlassian has built collision handling into the agent workflow: when an agent detects that someone else has edited the page mid-task, it pauses to avoid overwrites and previews its changes before publishing. This is a significant improvement over naive API-based automation that can silently overwrite concurrent edits.</p>
<h3>How does Atlassian&#8217;s agent platform compare to UiPath or Automation Anywhere for process automation?</h3>
<p>They serve different layers. UiPath and Automation Anywhere are purpose-built automation platforms with desktop automation (RPA), workflow orchestration, and process mining. Atlassian&#8217;s agent platform is a work management layer with agentic capabilities — it excels at software development, IT service management, and knowledge work automation within Atlassian&#8217;s product suite. The two approaches complement each other: an RPA bot can trigger Jira ticket creation via the MCP server, and a Rovo agent can invoke external automation workflows. They are not direct substitutes.</p>
<h2>Key Takeaways</h2>
<ul>
<li><strong>Atlassian is building the most complete &#8220;agents as teammates&#8221; model in enterprise software</strong> — with persistent identity, real-time presence, and full audit trails inside the tool where teams already work.</li>
<li><strong>The rebuilt MCP server (220+ tools, 15M+ daily calls, 25% fewer tokens)</strong> is a production-grade integration point that any MCP-compatible AI client can use, making it the widest open agent access surface among enterprise work management platforms.</li>
<li><strong>Multi-model routing</strong> (OpenAI GPT-6 Astra + Anthropic Claude + Google Gemini) means Rovo can optimize for quality, speed, and cost per task — no single-vendor lock-in on the reasoning layer.</li>
<li><strong>Bundled pricing with no current overage billing</strong> removes the experimentation barrier. Organizations can deploy agents across their entire Atlassian footprint without per-agent licensing.</li>
<li><strong>For RPA teams:</strong> Atlassian is not replacing your automation platform — it is providing a governed agent layer on top of your work management system. The MCP server is the bridge between agentic AI workflows and your Jira/Confluence data.</li>
<li><strong>Watch the gaps:</strong> Cross-vendor agent identity is unresolved, observability is dashboard-level (not production telemetry), and Rovo Work pricing and GA timing are unclear. Evaluate these before committing to production workloads.</li>
</ul>
<h2>References</h2>
<ol>
<li>Atlassian Blog — <a href="https://www.atlassian.com/blog/company-news/team26-europe-atlassian-mcp" target="_blank" rel="noopener nofollow">&#8220;Atlassian MCP brings more of your work within reach of any AI agent&#8221;</a> (October 2026)</li>
<li>VentureBeat — <a href="https://venturebeat.com/orchestration/atlassian-deepens-its-openai-partnership-with-a-spend-commitment-but-its-platform-stays-firmly-multi-model" target="_blank" rel="noopener nofollow">&#8220;Atlassian deepens its OpenAI partnership with a spend commitment&#8221;</a> (October 2026)</li>
<li>DEV Community — <a href="https://dev.to/mihai_leanzero/atlassian-team-26-europe-announcements-amp-mcp-1ek8" target="_blank" rel="noopener nofollow">&#8220;Atlassian Team &#8217;26 Europe announcements: AMP, MCP&#8221;</a> (October 2026)</li>
<li>Futurum Group — <a href="https://futurumgroup.com/insights/atlassian-fuses-the-agent-work-surface-workflow-and-control-plane-into-jira/" target="_blank" rel="noopener nofollow">&#8220;Atlassian Fuses the Agent Work Surface, Workflow, and Control Plane Into Jira&#8221;</a> (2026)</li>
<li>CCLeaks — <a href="https://ccleaks.com/news/atlassian-openai-rovo-mcp-partnership-oct-2026" target="_blank" rel="noopener nofollow">&#8220;Atlassian OpenAI Deal: How ChatGPT and Codex Reach Jira via MCP&#8221;</a> (October 2026)</li>
<li>Atlassian Blog — <a href="https://www.atlassian.com/blog/company-news/teamwork-collection-team-26" target="_blank" rel="noopener nofollow">&#8220;Teamwork Collection: Team &#8217;26&#8221;</a> (2026)</li>
<li>Atlassian Developer — <a href="https://developer.atlassian.com/cloud/rovo-mcp/" target="_blank" rel="noopener nofollow">&#8220;Atlassian Rovo MCP Overview&#8221;</a></li>
<li>GitHub — <a href="https://github.com/atlassian/atlassian-mcp-server" target="_blank" rel="noopener nofollow">atlassian/atlassian-mcp-server</a> (Apache-2.0)</li>
<li>eesel AI — <a href="https://www.eesel.ai/blog/atlassian-intelligence-and-rovo-pricing-explained" target="_blank" rel="noopener nofollow">&#8220;Atlassian Intelligence and Rovo pricing explained&#8221;</a> (2026)</li>
<li>Solutions Review — <a href="https://solutionsreview.com/ai-news-for-the-week-of-october-9-updates-from-honeycomb-io-ninjatech-ai-oracle-more/" target="_blank" rel="noopener nofollow">&#8220;AI News for the Week of October 9&#8221;</a> (October 2026)</li>
</ol>
<p><strong>Social Media Post (X/Twitter):</strong></p>
<p><em>Atlassian just turned Jira into an agent orchestration platform. 220+ MCP tools, real-time agent presence alongside humans, multi-model routing (GPT-6 Astra + Claude + Gemini), and bundled pricing. Here&#8217;s what RPA architects need to know: [link] #AgenticAI #RPA #Atlassian #MCP</em></p>
<p><strong>LinkedIn Post:</strong></p>
<p><em>Atlassian&#8217;s Team &#8217;26 Europe announcements are bigger than most coverage suggests.</em></p>
<p><em>They didn&#8217;t just add AI to Jira — they built a complete agent platform:</em></p>
<p><em>→ A rebuilt MCP server with 220+ tools handling 15M+ calls/day<br />
→ AMP (Agentic Multiplayer Protocol) — agents appear alongside humans in real-time<br />
→ Multi-model routing across OpenAI, Anthropic, and Google<br />
→ Agent accounts with persistent identity and full audit trails<br />
→ Rovo Work for multi-step tasks running hours in sandboxed environments<br />
→ All bundled into existing paid plans — no separate agent fee</em></p>
<p><em>For RPA teams: your Jira backlog just became an agent queue. Your Confluence pages just became agent workspaces. And the MCP server provides a single, standardized integration point that replaces dozens of custom connectors.</em></p>
<p><em>The full architecture breakdown: [link]</em></p>
<p><em>#AgenticAI #RPA #Automation #Atlassian #MCP #EnterpriseAI</em></p>
]]></content:encoded>
					
					<wfw:commentRss>https://rpabotsworld.com/atlassian-rovo-mcp-agentic-multiplayer-protocol-guide/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>Personal Agent Protocol (Poppy): The Complete Guide to Agentic Commerce&#8217;s Missing Identity Layer for RPA and Agentic AI Architects (2026)</title>
		<link>https://rpabotsworld.com/sierra-personal-agent-protocol-poppy-agentic-commerce-guide/</link>
					<comments>https://rpabotsworld.com/sierra-personal-agent-protocol-poppy-agentic-commerce-guide/#respond</comments>
		
		<dc:creator><![CDATA[Satish Prasad]]></dc:creator>
		<pubDate>Mon, 26 Oct 2026 05:09:01 +0000</pubDate>
				<category><![CDATA[Agentic AI & AI Automation]]></category>
		<category><![CDATA[AI Agents & Frameworks]]></category>
		<category><![CDATA[MCP]]></category>
		<guid isPermaLink="false">https://rpabotsworld.com/?p=32439</guid>

					<description><![CDATA[Personal Agent Protocol (Poppy) Draft 0.1 explained — OAuth flows, DPoP proofs, MCP integration, 40+ partners, and what RPA architects must prepare for now.]]></description>
										<content:encoded><![CDATA[<h1>Personal Agent Protocol (Poppy): The Complete Guide to Agentic Commerce&#8217;s Missing Identity Layer for RPA and Agentic AI Architects (2026)</h1>
<p>On September 20, 2026, Amazon blocked Meta&#8217;s personal AI agent, Muse, from completing purchases on its marketplace. Amazon&#8217;s objections were specific: Muse didn&#8217;t identify itself, didn&#8217;t ask permission, and appeared to capture and store customer credentials. Within sixteen days, Sierra and Meta published Draft 0.1 of the <strong>Personal Agent Protocol</strong> — nicknamed <strong>Poppy</strong> — an open, OAuth-based standard that gives AI agents a front door to businesses instead of forcing them to sneak through the browser like a human. By October 9, the partner list had swelled to over 40 organizations, including Bank of America, Visa, Mastercard, PayPal, Walmart, Target, Nordstrom, OpenAI, Okta, Cloudflare, and Zapier.</p>
<p>For RPA and agentic AI architects, the implications are immediate. The protocol that wins the identity layer will determine how every agent — from a simple invoice-processing bot to a multi-step procurement orchestrator — authenticates with external services. If you&#8217;re building automation that touches commerce, customer service, or any cross-organization workflow, Poppy is the specification you need to understand right now, even while it&#8217;s still in draft.</p>
<p>This guide breaks down the full Draft 0.1 specification, maps Poppy&#8217;s position in the emerging protocol stack alongside MCP, A2A, UCP, ACP, and Visa TAP, and gives you a concrete preparation checklist so your automation architecture is ready when the reference implementation ships.</p>
<h2>Table of Contents</h2>
<ol>
<li><a href="#what-is-poppy">What Is the Personal Agent Protocol (Poppy)?</a></li>
<li><a href="#why-it-matters-now">Why It Matters Now: The Amazon-Muse Conflict and the Identity Gap</a></li>
<li><a href="#the-protocol-stack">Where Poppy Fits in the Agentic Protocol Stack</a></li>
<li><a href="#five-building-blocks">The Five Building Blocks of Poppy</a></li>
<li><a href="#technical-deep-dive">Technical Deep Dive: OAuth, JWTs, DPoP, and Session Lifecycle</a></li>
<li><a href="#three-integration-modes">The Three Integration Modes: Website, APIs, and Company Agent</a></li>
<li><a href="#permission-model">The Permission Model: Guest, Read, and Write Access</a></li>
<li><a href="#conversation-protocol">The Conversation Protocol: Structured Agent-to-Business Dialogue</a></li>
<li><a href="#partner-ecosystem">The Partner Ecosystem: Who&#8217;s In and Who&#8217;s Missing</a></li>
<li><a href="#protocol-comparison">Protocol Comparison: Poppy vs. UCP vs. ACP vs. TAP vs. PACT</a></li>
<li><a href="#rpa-architect-implications">What This Means for RPA and Automation Architects</a></li>
<li><a href="#preparation-checklist">Preparation Checklist: 10 Steps to Get Ready</a></li>
<li><a href="#faq">Frequently Asked Questions</a></li>
<li><a href="#key-takeaways">Key Takeaways</a></li>
<li><a href="#references">References</a></li>
</ol>
<h2 id="what-is-poppy">What Is the Personal Agent Protocol (Poppy)?</h2>
<p>Poppy is an open protocol, licensed under Apache 2.0, that standardizes how a customer&#8217;s personal AI agent discovers a business, authenticates on the customer&#8217;s behalf, and completes tasks through the business&#8217;s website, APIs, or the company&#8217;s own AI agent. The name &#8220;Poppy&#8221; is the protocol&#8217;s internal codename, and the discovery file that businesses publish lives at <code>/.well-known/poppy.json</code>.</p>
<p>The core principle is a two-sided contract: <strong>customers decide what access to give their personal agents</strong>, and <strong>companies set the parameters for what those agents can do</strong>. The protocol doesn&#8217;t move money, process payments, or handle product discovery. It handles something more fundamental — identity, consent, and scope.</p>
<p>Think of it this way: MCP (<a href="https://rpabotsworld.com/a2a-joins-mcp-aaif-agentic-ai-architects-guide/">Model Context Protocol</a>) is the tool-access layer that lets agents call APIs. A2A is the agent-to-agent communication layer. Payment protocols like ACP and UCP handle checkout and settlement. Poppy sits alongside all of them as the <strong>identity and session layer</strong> — the part that answers &#8220;who is this agent, who does it represent, and what is it allowed to do?&#8221;</p>
<h3>What the Draft 0.1 Specification Covers</h3>
<p>The specification, published October 9, 2026 at <code>personalagentprotocol.org/docs/spec</code>, defines sixteen areas:</p>
<table>
<thead>
<tr>
<th>Area</th>
<th>What It Specifies</th>
</tr>
</thead>
<tbody>
<tr>
<td>Discovery</td>
<td>The <code>poppy.json</code> schema that businesses publish at a well-known URL</td>
</tr>
<tr>
<td>OAuth Server Verification</td>
<td>How agents verify a business&#8217;s OAuth issuer to prevent impersonation</td>
</tr>
<tr>
<td>Agent Identity</td>
<td>How agents prove who they are using <code>client_id</code> metadata and <code>private_key_jwt</code></td>
</tr>
<tr>
<td>Token Types</td>
<td>Short-lived DPoP-bound Session Tokens and long-lived Account Tokens</td>
</tr>
<tr>
<td>Session Lifecycle</td>
<td>Start, renew, sign in, reuse, and sign out flows</td>
</tr>
<tr>
<td>JWT Assertion Format</td>
<td>Claims structure for session start and renewal</td>
</tr>
<tr>
<td>DPoP Proof of Possession</td>
<td>Cryptographic binding of tokens to the presenting agent (RFC 9449)</td>
</tr>
<tr>
<td>Bearer Tokens for MCP</td>
<td>MCP-specific token handling since MCP cannot use DPoP</td>
</tr>
<tr>
<td>Three Sign-In Methods</td>
<td>Direct (authorization code + PKCE), Device, and Mediated</td>
</tr>
<tr>
<td>Permission Scopes</td>
<td>Standard <code>poppy:read</code> and <code>poppy:write</code> plus custom scopes</td>
</tr>
<tr>
<td>Account Token Refresh</td>
<td>Token rotation, narrowing scopes, secure storage requirements</td>
</tr>
<tr>
<td>Web Browsing Join</td>
<td>How an API session extends to browser-based interactions</td>
</tr>
<tr>
<td>API Integration</td>
<td>OpenAPI (DPoP) and MCP (Bearer) integration patterns</td>
</tr>
<tr>
<td>Conversation Protocol</td>
<td>Structured agent-to-business dialogue with handoff to humans</td>
</tr>
<tr>
<td>Extension Mechanism</td>
<td>How companies and agents add capabilities like payments</td>
</tr>
<tr>
<td>Token Error Reference</td>
<td>Standardized error codes and HTTP status codes</td>
</tr>
</tbody>
</table>
<p>This is not a white paper or a press release with aspirational language — it&#8217;s a real specification with JWT claim formats, DPoP proof structures, and HTTP error codes. That level of detail is what separates Poppy from the wave of &#8220;we&#8217;ll publish a spec eventually&#8221; announcements that have characterized the agentic commerce space.</p>
<h2 id="why-it-matters-now">Why It Matters Now: The Amazon-Muse Conflict and the Identity Gap</h2>
<p>The catalyst for Poppy was a real-world incident that exposed the absence of an identity standard for personal AI agents. Here&#8217;s the timeline:</p>
<p><strong>September 8, 2026:</strong> Meta releases Muse, a personal AI agent that can browse the web, make purchases, and manage tasks across email, calendar, and commerce services. Payments run through Stripe&#8217;s Link, which creates a one-time virtual card for each transaction. Muse runs in a secure cloud virtual machine with its own browser and uses a monitoring agent called Sentinel to approve outbound traffic.</p>
<p><strong>September 15, 2026:</strong> Muse hits #1 in Apple&#8217;s US App Store.</p>
<p><strong>September 20, 2026:</strong> Amazon blocks Muse. Amazon&#8217;s stated objections, <a href="https://ppc.land/amazon-blocks-metas-muse-ai-agent-from-shopping-on-its-site/" target="_blank" rel="noopener nofollow">as reported by GeekWire</a>, include that Muse doesn&#8217;t identify itself as an agent when browsing, that Meta didn&#8217;t get Amazon&#8217;s consent, and that the agent appears to capture and store customer credentials. Amazon&#8217;s spokesperson argued that &#8220;third-party shopping apps should operate openly and respect service provider decisions about whether or not to participate.&#8221;</p>
<p><strong>October 6, 2026:</strong> Sierra and Meta announce the Personal Agent Protocol with initial partners Genesys, Instinct, NiCE, Rocket, Shopify, Stripe, and Walmart.</p>
<p><strong>October 9, 2026:</strong> Sierra publishes Draft 0.1 of the specification, along with 35 new design partners including Bank of America, Visa, Mastercard, OpenAI, Okta, Wells Fargo, Target, Nordstrom, United Airlines, and Zapier.</p>
<p>The pattern is clear: the &#8220;chaos&#8221; Bret Taylor described — where agents log in as users and click through web pages designed for humans — had real consequences. Amazon&#8217;s decision to block Muse wasn&#8217;t just a business dispute; it was a structural problem. Without a standard for agent identity, every interaction between an agent and a business is either a hack (pretending to be the user) or a bespoke integration (building a custom API for each agent provider). Neither scales.</p>
<h3>Why RPA Architects Should Pay Attention</h3>
<p>If you&#8217;ve built RPA automations that interact with external websites or services, you&#8217;ve lived this problem. Your bots log in with stored credentials, navigate UIs designed for humans, and break when those UIs change. The identity gap that Poppy addresses isn&#8217;t unique to consumer shopping agents — it&#8217;s the same gap that makes enterprise automation fragile. A standard identity layer could transform how automation interacts with SaaS platforms, vendor portals, and partner systems.</p>
<h2 id="the-protocol-stack">Where Poppy Fits in the Agentic Protocol Stack</h2>
<p>The agentic AI ecosystem is rapidly developing a layered protocol stack. Understanding where each protocol sits prevents the confusion of treating them as competitors when they&#8217;re mostly complementary layers.</p>
<table>
<thead>
<tr>
<th>Layer</th>
<th>Protocol(s)</th>
<th>What It Handles</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Identity &amp; Session</strong></td>
<td>Poppy (PAP), Visa TAP, PACT</td>
<td>Who is this agent? Who does it represent? What can it do?</td>
</tr>
<tr>
<td><strong>Tool Access</strong></td>
<td>MCP (Model Context Protocol)</td>
<td>How agents call tools, read resources, and access data</td>
</tr>
<tr>
<td><strong>Agent-to-Agent</strong></td>
<td>A2A (Agent-to-Agent Protocol)</td>
<td>How agents delegate tasks across vendors and organizations</td>
</tr>
<tr>
<td><strong>Commerce Journey</strong></td>
<td>UCP (Universal Commerce Protocol)</td>
<td>Discovery → checkout → post-purchase across the full buyer flow</td>
</tr>
<tr>
<td><strong>Payment Authorization</strong></td>
<td>AP2, ACP, Mastercard Agent Pay</td>
<td>User consent, payment tokens, and settlement</td>
</tr>
<tr>
<td><strong>Settlement</strong></td>
<td>Existing payment rails (Visa/MC/Stripe/PayPal)</td>
<td>Moving money between accounts</td>
</tr>
</tbody>
</table>
<p>Poppy explicitly positions itself as complementary to all of these. From the <a href="https://sierra.ai/blog/poppy" target="_blank" rel="noopener nofollow">Sierra blog post</a>: &#8220;Universal Commerce Protocol focuses on shopping — from product discovery through checkout and order management. Poppy covers how an agent identifies itself, gets permission, and works with a company on any task, from buying a sweater to applying for a mortgage.&#8221;</p>
<p>The key insight for architects is the word &#8220;any task.&#8221; UCP is scoped to commerce. Poppy is scoped to identity across all interactions — commerce, customer service, insurance claims, mortgage applications, medical appointments. That breadth is what makes it relevant beyond shopping bots.</p>
<h2 id="five-building-blocks">The Five Building Blocks of Poppy</h2>
<h3>1. Discovery: The poppy.json File</h3>
<p>Every participating business publishes a single JSON file at <code>https://{domain}/.well-known/poppy.json</code>. This file tells agents how to interact with the company, how customers sign in, and what the agent can do. The specification requires HTTPS and allows redirects only to HTTPS URLs.</p>
<p>The top-level fields include:</p>
<ul>
<li><code>protocol_version</code>: Currently <code>"0.1"</code></li>
<li><code>organization</code>: Company name and domain (must match the request host)</li>
<li><code>auth</code>: OAuth issuer, supported sign-in methods (direct, device, mediated), and custom scopes</li>
<li><code>agent</code>: Conversation protocol endpoints</li>
<li><code>apis</code>: Array of API definitions (OpenAPI or MCP) with URLs and descriptions</li>
<li><code>web</code>: Browser session endpoint for web browsing integration</li>
<li><code>extensions</code>: Optional capability extensions</li>
</ul>
<p>For RPA architects, this is analogous to a <code>robots.txt</code> for agents — but instead of &#8220;don&#8217;t crawl these pages,&#8221; it says &#8220;here&#8217;s how to interact with us properly.&#8221; The single-file approach makes adoption straightforward: publish one JSON file, and every compliant agent knows how to find you.</p>
<h3>2. Sessions: One Identity Across Channels</h3>
<p>When an agent starts interacting with a business, it creates a session. The session can begin as a guest (no sign-in required) and upgrade to an authenticated session when the user signs in. Crucially, the same session token works across the business&#8217;s APIs, website, and company agent. A question asked before sign-in and an order placed after sign-in belong to the same visit.</p>
<p>Sessions use short-lived OAuth access tokens, DPoP-bound (Demonstration of Proof-of-Possession, RFC 9449), which means the token is cryptographically tied to the specific agent presenting it. Stolen tokens can&#8217;t be replayed by a different client.</p>
<h3>3. Sign-In: OAuth with Three Methods</h3>
<p>Poppy supports three sign-in methods, each suited to different interaction contexts:</p>
<table>
<thead>
<tr>
<th>Method</th>
<th>How It Works</th>
<th>Best For</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Direct</strong></td>
<td>Standard OAuth authorization code flow with PKCE (RFC 7636). The user signs in via browser.</td>
<td>First-time setup, high-trust actions</td>
</tr>
<tr>
<td><strong>Device</strong></td>
<td>Device authorization grant (RFC 8628). Agent shows a code; user opens a URL on any device.</td>
<td>Headless agents, voice interfaces</td>
</tr>
<tr>
<td><strong>Mediated</strong></td>
<td>Agent posts credentials to a specific endpoint. Fields marked <code>secret</code> must never be logged.</td>
<td>Repeat interactions where credentials are pre-stored</td>
</tr>
</tbody>
</table>
<p>The specification is explicit that Direct and Device Sign-In must only be used when the user personally completes the sign-in. This prevents agents from using stored credentials to silently authenticate — precisely the behavior Amazon objected to with Muse.</p>
<h3>4. Getting the Job Done: Three Integration Modes</h3>
<p>Once authenticated, the agent interacts through one or more channels chosen by the business (detailed in the section below).</p>
<h3>5. Extensions: A Future-Proof Architecture</h3>
<p>The specification includes a formal extension mechanism. Companies and agents list supported extensions in their discovery files and metadata. Spec-defined extensions use plain names; third-party extensions use a domain the author controls (e.g., <code>example.com/gift-wrap</code>). This is how payments, push notifications, and industry-specific capabilities will be added without breaking the core protocol.</p>
<h2 id="technical-deep-dive">Technical Deep Dive: OAuth, JWTs, DPoP, and Session Lifecycle</h2>
<p>This section is for architects who need to understand the wire-level protocol. If you&#8217;re looking for the business implications, skip to the <a href="#rpa-architect-implications">RPA Architect Implications</a> section.</p>
<h3>Agent Identity Verification</h3>
<p>Every agent has a <code>client_id</code> that is an HTTPS URL pointing to its metadata document. This document follows the OAuth Client ID Metadata Document draft and includes:</p>
<ul>
<li><code>client_name</code> and <code>logo_uri</code> for display</li>
<li><code>jwks_uri</code> pointing to the agent&#8217;s public keys (HTTPS, same domain as <code>client_id</code>)</li>
<li><code>redirect_uris</code> for OAuth callbacks</li>
<li><code>token_endpoint_auth_method</code>: always <code>private_key_jwt</code></li>
</ul>
<p>Client authentication uses RFC 7523 (<code>private_key_jwt</code>), signed with a key from the agent&#8217;s published JWKS. This means every API call from an agent carries a cryptographic proof of the agent&#8217;s identity — a direct answer to Amazon&#8217;s complaint that Muse didn&#8217;t identify itself.</p>
<h3>Session Token Lifecycle</h3>
<p>The lifecycle follows five states:</p>
<ol>
<li><strong>Start a Session:</strong> The agent requests a signed-out Session Token from the company&#8217;s token endpoint using a JWT bearer grant (RFC 7523). The JWT assertion includes the agent&#8217;s <code>client_id</code> as <code>iss</code>, the user&#8217;s opaque company-specific ID as <code>sub</code>, and a 128-bit random <code>jti</code> to prevent replay attacks.</li>
<li><strong>Renew:</strong> Repeat the request with the existing <code>session_id</code>. Expired sessions get fresh tokens without losing context.</li>
<li><strong>Sign In:</strong> Use one of the three sign-in methods. This returns an Account Token (long-lived refresh token) and a signed-in Session Token.</li>
<li><strong>Reuse:</strong> Exchange the Account Token for signed-in tokens in new or existing sessions.</li>
<li><strong>Sign Out:</strong> Revoke the Account Token at the revocation endpoint (RFC 7009). Sessions continue in signed-out state.</li>
</ol>
<h3>DPoP Proof of Possession</h3>
<p>All non-MCP API calls require DPoP proofs (RFC 9449). Each proof is a JWT with:</p>
<ul>
<li><code>typ: dpop+jwt</code></li>
<li><code>alg: ES256</code> (ECDSA with P-256)</li>
<li>The agent&#8217;s public key in the <code>jwk</code> header</li>
<li>Claims: <code>jti</code> (unique), <code>htm</code> (HTTP method), <code>htu</code> (URL without query), <code>iat</code>, and <code>ath</code> (SHA-256 hash of the Session Token)</li>
</ul>
<p>The company verifies the signature, checks key binding against the token, validates method and URL match, confirms <code>iat</code> is within its acceptance window (typically 60 seconds), and rejects reused <code>jti</code> values. This makes token theft largely useless — a stolen token can&#8217;t be used from a different agent.</p>
<h3>MCP-Specific Token Handling</h3>
<p>MCP cannot use DPoP (its transport doesn&#8217;t support the required headers), so MCP access uses standard Bearer tokens. The specification requires that each Bearer token be scoped to a single MCP server using RFC 8707 resource indicators. Companies <strong>must</strong> accept Bearer tokens only on MCP APIs and <strong>must</strong> require DPoP everywhere else. This is a pragmatic compromise that keeps MCP integration simple while maintaining security on all other endpoints.</p>
<h2 id="three-integration-modes">The Three Integration Modes: Website, APIs, and Company Agent</h2>
<p>Businesses choose which channels to offer agents. Most will offer more than one, and the choice shapes how automation architects design their integrations.</p>
<h3>Mode 1: Website (Browser-Based)</h3>
<p>The agent navigates the business&#8217;s existing web pages using a browser it controls. The session is joined to the browser via a special JWT assertion (type <code>poppy-browser+jwt</code>) delivered as a form POST. The company sets a <code>Secure</code>, <code>HttpOnly</code>, <code>SameSite</code> cookie that reflects the session state and applies the user&#8217;s permissions to each page.</p>
<p><strong>For RPA architects:</strong> This is the closest mode to traditional RPA. Your bot navigates web pages, but with a crucial difference — it identifies itself as an agent, carries authenticated session state, and operates within permissions the user explicitly granted. No more credential scraping or pretending to be a human.</p>
<h3>Mode 2: APIs (OpenAPI and MCP)</h3>
<p>The structured integration path. Businesses expose OpenAPI endpoints (using DPoP tokens) or <a href="https://rpabotsworld.com/a2a-joins-mcp-aaif-agentic-ai-architects-guide/">MCP servers</a> (using Bearer tokens). Each API keeps its own schema, and MCP servers must publish RFC 9728 protected resource metadata.</p>
<p>Standard HTTP error handling applies: <code>401 invalid_token</code> for expired or invalid tokens, <code>403 sign_in_required</code> when authentication is needed, and <code>403 insufficient_scope</code> when the agent&#8217;s permissions are too narrow (with the required scope returned in the response).</p>
<p><strong>For RPA architects:</strong> This is the path you should plan for. As businesses publish Poppy-compliant APIs, the value of UI-based automation drops. Instead of maintaining fragile selectors on a checkout page, you call a structured API with proper error codes and scope management.</p>
<h3>Mode 3: Company Agent (Conversational)</h3>
<p>The most novel integration mode. The personal agent hands off complex, dialogue-heavy tasks — warranty claims, mortgage applications, insurance disputes — to the business&#8217;s own AI agent. The conversation protocol supports text, structured data, status tracking, streaming via SSE, human handoff, and &#8220;direct conversations&#8221; (sub-conversations that lock the parent while open).</p>
<p>Each message carries a <code>sender</code> field (<code>agent</code> or <code>human</code>), so both sides always know whether they&#8217;re talking to an AI or a person. The <code>context</code> field carries locale, timezone, and user availability — critical for customer-service interactions.</p>
<p><strong>For RPA architects:</strong> Agent-to-agent conversations are the next frontier beyond API integrations. If you&#8217;re building customer-facing automations — chatbots, support agents, or service desk workflows — this protocol mode defines how your agent will interact with your customers&#8217; personal agents.</p>
<h2 id="permission-model">The Permission Model: Guest, Read, and Write Access</h2>
<p>Poppy defines a tiered permission model that maps cleanly to enterprise access-control patterns:</p>
<table>
<thead>
<tr>
<th>Tier</th>
<th>Authentication</th>
<th>What the Agent Can Do</th>
<th>Example</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Blocked</strong></td>
<td>N/A</td>
<td>Nothing — the business doesn&#8217;t participate</td>
<td>A business that opts out entirely</td>
</tr>
<tr>
<td><strong>Guest</strong></td>
<td>Agent identifies itself, no user sign-in</td>
<td>Read public information</td>
<td>Check stock, read return policy, view store hours</td>
</tr>
<tr>
<td><strong>Signed-in, Read</strong></td>
<td>OAuth sign-in, <code>poppy:read</code> scope</td>
<td>View account information</td>
<td>Check order status, view balances, see appointments</td>
</tr>
<tr>
<td><strong>Signed-in, Write</strong></td>
<td>OAuth sign-in, <code>poppy:write</code> scope</td>
<td>Make changes to the account</td>
<td>Start a return, rebook a flight, update payment method</td>
</tr>
<tr>
<td><strong>Custom Scopes</strong></td>
<td>OAuth sign-in, company-defined scopes</td>
<td>Fine-grained access to specific actions</td>
<td><code>addresses</code> scope for address changes only</td>
</tr>
</tbody>
</table>
<p>The specification enforces several important constraints:</p>
<ul>
<li>Scopes are independent — granting <code>poppy:write</code> does not automatically include <code>poppy:read</code></li>
<li>Users may approve fewer scopes than requested, so agents must check the returned <code>scope</code> value</li>
<li>Re-signing in with additional scopes replaces the Account Token (the old one should be revoked)</li>
<li>The <code>poppy:</code> prefix is reserved for protocol-defined scopes</li>
</ul>
<p>For enterprise scenarios, the <a href="https://www.beri.net/article/meta-sierra-personal-agent-protocol-oauth-guest-read-write-access-vs-ucp-acp-trusted-agent-protocol-retail-banks" target="_blank" rel="noopener nofollow">beri.net analysis</a> recommends a four-tier permission matrix: guest read (public data), signed-in read (account data), signed-in write reversible (returns, rescheduling), and signed-in write irreversible (purchases, transfers) — with the irreversible tier kept closed until payments arrive in a future spec version.</p>
<h2 id="conversation-protocol">The Conversation Protocol: Structured Agent-to-Business Dialogue</h2>
<p>The conversation protocol is one of Poppy&#8217;s most architecturally significant components. It defines a structured, bidirectional dialogue format that goes well beyond simple request-response patterns.</p>
<h3>Endpoints</h3>
<ul>
<li><code>POST {endpoint}</code>: Start a new conversation</li>
<li><code>POST {endpoint}/{id}/messages</code>: Send a message (202 Accepted, or 201 Created for the first message)</li>
<li><code>GET {endpoint}/{id}/events</code>: Read events with long-polling (<code>cursor</code> and <code>wait</code>) or SSE</li>
<li><code>POST {endpoint}/{id}/handoff</code>: Request a human agent</li>
<li><code>POST {endpoint}/{id}/close</code>: Close the conversation</li>
</ul>
<h3>Message Structure</h3>
<p>Each message carries:</p>
<ul>
<li><code>id</code>: Chosen by the sender, unique per user. Retries with identical content are idempotent.</li>
<li><code>sender</code>: Either <code>agent</code> or <code>human</code> — both sides always declare who&#8217;s speaking</li>
<li><code>text</code>: Natural language content</li>
<li><code>data</code>: Structured payload for machine-readable details</li>
<li><code>context</code>: Locale (RFC 5646), timezone (IANA), user availability</li>
</ul>
<h3>Event Types and Status Tracking</h3>
<p>Conversations emit typed events: <code>message</code>, <code>state</code> (working/idle/queued/closed), <code>authorization</code> (sign-in triggered mid-conversation), <code>user_requested</code> (the business wants the user involved), <code>direct_opened</code>, and <code>direct_closed</code>. Agents skip unknown event types, which enables forward-compatible extensions.</p>
<p>The &#8220;direct conversation&#8221; pattern is particularly relevant for complex workflows: a personal agent can open a sub-conversation with a business agent for a specific sub-task (like a warranty claim within a broader support interaction), and the parent conversation is locked until the sub-conversation closes. Only one direct conversation can be open per parent at a time.</p>
<h2 id="partner-ecosystem">The Partner Ecosystem: Who&#8217;s In and Who&#8217;s Missing</h2>
<p>The coalition behind Poppy expanded dramatically between the October 6 announcement and the October 9 draft publication. The partner list now spans financial services, retail, insurance, travel, technology, and media.</p>
<h3>Founding Partners (October 6, 2026)</h3>
<p>Meta, Sierra, Genesys, Instinct, NiCE, Rocket, Shopify, Stripe, Walmart</p>
<h3>Design Partners (October 9, 2026)</h3>
<p>Adyen, Atomic, Bank of America, BBVA, Chime, Cigna, Cloudflare, Comcast, DIRECTV, ElevenLabs, FOX, Gap Inc., GEICO, Hertz, Insurify, Klaviyo, Liberty Mutual, Mastercard, Nordstrom, Notion, Okta, OneSignal, <strong>OpenAI</strong>, PayPal, Plaid, SiriusXM, Synchrony, Target, United Airlines, Venmo, Visa, Wells Fargo, Zapier, Zendesk, 1Password</p>
<h3>Notable Additions and Overlaps</h3>
<p><strong>OpenAI joining as a design partner</strong> is significant. The initial announcement conspicuously excluded OpenAI and Anthropic — Bret Taylor (Sierra co-founder and OpenAI board chair) had said he expected them to join. OpenAI&#8217;s inclusion in the October 9 design partner list signals convergence rather than fragmentation.</p>
<p><strong>Cross-protocol participation</strong> is common. Shopify, Stripe, and Walmart sit on UCP&#8217;s governing council, Visa backs both TAP and the Poppy coalition, and Stripe co-developed ACP with OpenAI. This hedging suggests the industry expects a layered stack rather than a winner-take-all protocol war.</p>
<h3>Who&#8217;s Not Listed</h3>
<p>As of October 12, 2026:</p>
<ul>
<li><strong>Anthropic</strong> is absent. As the creator of MCP, their absence from the identity layer is notable.</li>
<li><strong>Amazon</strong> is absent — unsurprising given the Muse conflict, but significant since Amazon controls roughly 40% of US e-commerce.</li>
<li><strong>Google</strong> is absent. Google backs UCP and A2A, and its Gemini agent platform is a potential competitor.</li>
<li><strong>No European-headquartered retailers or banks</strong> beyond Stripe (which is dual-headquartered in Dublin and San Francisco).</li>
</ul>
<h2 id="protocol-comparison">Protocol Comparison: Poppy vs. UCP vs. ACP vs. TAP vs. PACT</h2>
<p>The agentic commerce protocol landscape is crowded. This comparison table maps each protocol&#8217;s scope, backers, and current status.</p>
<table>
<thead>
<tr>
<th>Protocol</th>
<th>Primary Backers</th>
<th>Layer</th>
<th>Scope</th>
<th>Spec Status (Oct 2026)</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Poppy (PAP)</strong></td>
<td>Sierra, Meta, 40+ partners</td>
<td>Identity &amp; Session</td>
<td>Agent authentication, consent, session management across any business task</td>
<td>Draft 0.1 published Oct 9</td>
</tr>
<tr>
<td><strong>UCP</strong></td>
<td>Google, Shopify, Walmart, Target, Visa, Stripe, PayPal</td>
<td>Commerce Journey</td>
<td>Full purchase lifecycle: discovery → checkout → post-purchase</td>
<td>Available, announced NRF 2026</td>
</tr>
<tr>
<td><strong>ACP</strong></td>
<td>OpenAI, Stripe</td>
<td>Checkout</td>
<td>Shared Payment Token for agent-initiated checkout (one-use, time-bounded)</td>
<td>Spec published; ChatGPT Instant Checkout retired Mar 2026</td>
</tr>
<tr>
<td><strong>Visa TAP</strong></td>
<td>Visa, Cloudflare, Microsoft, Stripe, Shopify</td>
<td>Agent Identity + Payment</td>
<td>HTTP Message Signatures (RFC 9421) for agent identity and payment authorization</td>
<td>Launched Oct 2025; connected to ChatGPT Jun 2026</td>
</tr>
<tr>
<td><strong>AP2</strong></td>
<td>Google, PayPal</td>
<td>Payment Authorization</td>
<td>Signed user mandates with merchant ID, amount, expiry</td>
<td>Production-ready</td>
</tr>
<tr>
<td><strong>PACT</strong></td>
<td>Decagon, Instinct</td>
<td>Consent &amp; Trust</td>
<td>A2A Agent Cards for discovery, scoped agent JWTs, signed receipts</td>
<td>Open-source spec at openpactprotocol.org</td>
</tr>
<tr>
<td><strong>MCP</strong></td>
<td>Anthropic → AAIF (Linux Foundation)</td>
<td>Tool Access</td>
<td>Agent-to-tool data plane: tools, resources, prompts</td>
<td>Production, broadest adoption</td>
</tr>
<tr>
<td><strong>A2A</strong></td>
<td>Google → AAIF (Linux Foundation)</td>
<td>Agent-to-Agent</td>
<td>Cross-vendor task delegation via task envelopes</td>
<td>v1.0 published; 150+ orgs in production</td>
</tr>
</tbody>
</table>
<h3>The Complementarity Argument</h3>
<p>Sierra positions Poppy as complementary to all of these. The argument is that MCP handles tool access, A2A handles agent-to-agent delegation, UCP handles the commerce journey, payment protocols handle settlement, and Poppy handles identity across all of them. The <a href="https://forkast.news/the-personal-agent-protocol-is-not-a-standard-war-its-the-missing-identity-layer/" target="_blank" rel="noopener nofollow">Forkast analysis</a> frames it as: &#8220;the gateway becomes a protocol multiplexer that must support PAP, MCP, A2A, Visa TAP, and possibly ACP.&#8221;</p>
<h3>The Overlap Reality</h3>
<p>Despite the complementarity framing, there is real overlap in the identity space. Visa TAP also handles agent identity via HTTP Message Signatures. UCP could absorb the identity layer. PACT provides consent and trust with actual open-source code. The market hasn&#8217;t decided whether identity will be a separate protocol or a feature of the commerce and payment protocols.</p>
<h2 id="rpa-architect-implications">What This Means for RPA and Automation Architects</h2>
<h3>1. The Shift From UI Automation to Authenticated API Access</h3>
<p>Poppy&#8217;s three integration modes — website, APIs, and company agent — map directly to the evolution every RPA team is navigating. Today&#8217;s UI-based bots are the equivalent of Poppy&#8217;s &#8220;website&#8221; mode, but with none of the identity infrastructure. As businesses publish <code>poppy.json</code> files and expose MCP or OpenAPI endpoints, the value proposition of screen-scraping drops. <a href="https://rpabotsworld.com/microsoft-copilot-autopilot-code-home-agentic-enterprise-guide/">Copilot-style</a> agent architectures that already use structured APIs will have a natural migration path.</p>
<h3>2. Agent-to-Agent Conversations Are a New Integration Pattern</h3>
<p>The conversation protocol is something RPA architectures don&#8217;t typically model. When a personal agent negotiates a mortgage pre-approval with Rocket&#8217;s agent (as demonstrated at Sierra Summit), that&#8217;s not an API call or a web page navigation — it&#8217;s a stateful, multi-turn dialogue with structured data exchange, human handoff capability, and session persistence. If you&#8217;re building customer-service or sales automations, this pattern will become a requirement, not an option.</p>
<h3>3. Credential Management Gets Standardized</h3>
<p>One of the most painful aspects of enterprise RPA is credential management — storing passwords, rotating tokens, handling MFA. Poppy&#8217;s OAuth-based model replaces this with standard token flows. The specification requires that Account Tokens be stored as secrets and kept out of model context, messages, logs, and URLs. For regulated industries, this is a significant compliance improvement over the status quo of bots holding user passwords in credential vaults.</p>
<h3>4. Permission Scoping Changes the Automation Design Pattern</h3>
<p>In traditional RPA, a bot that logs in as a user inherits all of that user&#8217;s permissions. With Poppy, permissions are explicitly scoped — <code>poppy:read</code>, <code>poppy:write</code>, or custom scopes. This means automation architects need to design for <em>minimum required scope</em> rather than <em>full user access</em>. The specification even enforces this: users may approve fewer scopes than requested, so agents must handle narrowed permissions gracefully.</p>
<h3>5. Your Automation Platform Is About to Become a &#8220;Company Agent&#8221;</h3>
<p>If you build customer-facing automations — chatbots, IVR agents, support desk bots — you&#8217;re building what Poppy calls the &#8220;company agent.&#8221; Personal agents will arrive at your front door expecting to converse through the Poppy conversation protocol. Your automation platform needs to understand sessions, handle the <code>sender</code> field (is it talking to an agent or a human?), support direct conversations for sub-tasks, and gracefully hand off to human agents when needed.</p>
<h2 id="preparation-checklist">Preparation Checklist: 10 Steps to Get Ready</h2>
<p>Poppy&#8217;s reference implementation hasn&#8217;t shipped yet, but the draft specification is detailed enough to start preparing. Here&#8217;s what to do now:</p>
<ol>
<li><strong>Audit your external-facing APIs.</strong> Inventory every API and webhook your automation interacts with. Which ones could expose a <code>poppy.json</code> file? Which currently require screen-scraping that a Poppy integration would replace?</li>
<li><strong>Map your current OAuth infrastructure.</strong> Poppy sits on OAuth 2.0, JWTs, and DPoP. Assess whether your identity stack (Okta, Azure AD, Auth0) supports DPoP (RFC 9449) and device authorization grants (RFC 8628).</li>
<li><strong>Inventory credential vault dependencies.</strong> List every automation that stores user credentials. These are candidates for migration to Poppy&#8217;s token-based model. Prioritize the ones where credential rotation is already a pain point.</li>
<li><strong>Define your permission tiers.</strong> Use the four-tier matrix: guest read, signed-in read, signed-in write (reversible), and signed-in write (irreversible). Map your current automation actions into these tiers.</li>
<li><strong>Evaluate your MCP readiness.</strong> Poppy uses MCP as one of its API integration modes, with Bearer tokens scoped per server via RFC 8707 resource indicators. If you&#8217;ve already invested in <a href="https://rpabotsworld.com/a2a-joins-mcp-aaif-agentic-ai-architects-guide/">MCP infrastructure</a>, you&#8217;re ahead.</li>
<li><strong>Design for scope narrowing.</strong> Build your automations to handle getting fewer permissions than requested. This means checking the <code>scope</code> value in every token response and degrading gracefully.</li>
<li><strong>Plan your company agent architecture.</strong> If you run customer-facing bots, sketch how they&#8217;d expose a Poppy conversation endpoint: message handling, status tracking, streaming, and human handoff.</li>
<li><strong>Log agent identity separately.</strong> Start tagging interactions from automated agents differently from human users in your analytics. When Poppy goes live, you&#8217;ll want historical data on agent traffic patterns.</li>
<li><strong>Watch the reference implementation.</strong> Sierra has promised it &#8220;within about a month&#8221; of October 9. This is what you&#8217;ll build against.</li>
<li><strong>Track the coalition.</strong> OpenAI joined as a design partner. If Anthropic and Google follow, Poppy becomes the de facto identity layer. If they don&#8217;t, plan for supporting multiple identity protocols.</li>
</ol>
<h2 id="faq">Frequently Asked Questions</h2>
<h3>Does the Personal Agent Protocol handle payments?</h3>
<p>No. Draft 0.1 explicitly excludes payments. Sierra lists payment extensions as future work alongside push notifications and finer-grained permissions. For now, Poppy handles identity and session management; payment settlement stays with existing protocols like ACP, AP2, and the standard card networks. The extension mechanism is designed so industries can add payment capabilities without breaking the core protocol.</p>
<h3>How does Poppy relate to MCP (Model Context Protocol)?</h3>
<p>MCP is one of the API integration modes within Poppy. When a business publishes MCP server endpoints in its <code>poppy.json</code>, agents connect to those servers using Bearer tokens (since MCP&#8217;s transport doesn&#8217;t support DPoP). Poppy adds the identity and session layer that MCP itself doesn&#8217;t provide — authentication, consent, and cross-channel session persistence. They&#8217;re designed to work together, not replace each other.</p>
<h3>Can a business block all AI agents under Poppy?</h3>
<p>Yes. Poppy explicitly supports a &#8220;Blocked&#8221; tier where businesses choose not to participate at all. Businesses also have fine-grained controls: they can operate allowlists or blocklists by agent <code>client_id</code>, rate-limit specific agents, require agent registration before access, and revoke access at any time. The protocol is designed to give businesses control, not to force them open.</p>
<h3>Is Poppy only for consumer shopping?</h3>
<p>No, and this is a common misconception. While the Amazon-Muse commerce dispute triggered Poppy&#8217;s development, the protocol is designed for any agent-to-business interaction. Sierra&#8217;s blog gives examples ranging from retail returns to mortgage pre-approvals to flight rebookings. The partner list includes banks (Bank of America, Wells Fargo, BBVA), insurance companies (GEICO, Liberty Mutual, Cigna), airlines (United Airlines), and enterprise SaaS providers (Zendesk, Notion, 1Password). Any scenario where a personal agent needs to authenticate with a business and complete a task is in scope.</p>
<h3>What&#8217;s the difference between Poppy and Visa&#8217;s Trusted Agent Protocol (TAP)?</h3>
<p>TAP uses HTTP Message Signatures (RFC 9421) to sign agent identity and payment data into HTTP requests. It&#8217;s focused on the moment of payment — verifying that a specific agent is authorized to transact on a user&#8217;s behalf. Poppy covers the broader session — discovery, authentication, ongoing interaction across multiple channels. They could coexist, with Poppy handling the session and TAP handling the payment authorization within that session. Notably, Visa is a design partner for both protocols.</p>
<h2 id="key-takeaways">Key Takeaways</h2>
<ul>
<li><strong>Poppy (Personal Agent Protocol) is the first draft specification for an agent identity layer</strong> — it standardizes how AI agents discover businesses, authenticate on behalf of users, and operate within explicitly granted permissions.</li>
<li><strong>Draft 0.1 is technically detailed</strong>, specifying OAuth flows, JWT assertion formats, DPoP proof-of-possession, MCP Bearer token handling, a full conversation protocol, and an extension mechanism. This is a buildable specification, not a position paper.</li>
<li><strong>The coalition is massive and growing fast</strong> — from 9 founding partners to 40+ in three days, including major banks (Bank of America, Wells Fargo), payment networks (Visa, Mastercard, PayPal), retailers (Walmart, Target, Nordstrom), and notably OpenAI.</li>
<li><strong>Poppy sits alongside MCP and A2A, not above them.</strong> MCP provides tool access, A2A provides agent-to-agent communication, and Poppy provides the identity and session layer that answers &#8220;who is this agent and what can it do?&#8221;</li>
<li><strong>For RPA architects, the implications are structural</strong> — from UI automation shifting to authenticated API access, to credential management getting standardized, to your customer-facing bots needing to support the Poppy conversation protocol as incoming personal agents become a channel.</li>
<li><strong>Payments are not in Draft 0.1</strong>, but the extension mechanism is designed to add them. The protocol&#8217;s value is in the identity layer, not the settlement layer.</li>
<li><strong>Start preparing now</strong> — audit your APIs, map your OAuth infrastructure, evaluate MCP readiness, and design for scoped permissions. The reference implementation is expected within a month.</li>
</ul>
<h2 id="references">References</h2>
<ol>
<li>Sierra. &#8220;Poppy: Personal Agent Protocol.&#8221; October 9, 2026. <a href="https://sierra.ai/blog/poppy" target="_blank" rel="noopener nofollow">https://sierra.ai/blog/poppy</a></li>
<li>Personal Agent Protocol Specification, Draft 0.1. <a href="https://personalagentprotocol.org/docs/spec" target="_blank" rel="noopener nofollow">https://personalagentprotocol.org/docs/spec</a></li>
<li>Personal Agent Protocol Overview. <a href="https://personalagentprotocol.org/docs/overview" target="_blank" rel="noopener nofollow">https://personalagentprotocol.org/docs/overview</a></li>
<li>Yahoo Tech. &#8220;Sierra and Meta&#8217;s Personal Agent Protocol Gives Commerce a Common Front Door.&#8221; <a href="https://tech.yahoo.com/ai/meta-ai/articles/sierra-meta-personal-agent-protocol-024451270.html" target="_blank" rel="noopener nofollow">https://tech.yahoo.com/ai/meta-ai/articles/sierra-meta-personal-agent-protocol-024451270.html</a></li>
<li>Forkast. &#8220;The Personal Agent Protocol Is Not a Standard War — It&#8217;s the Missing Identity Layer.&#8221; <a href="https://forkast.news/the-personal-agent-protocol-is-not-a-standard-war-its-the-missing-identity-layer/" target="_blank" rel="noopener nofollow">https://forkast.news/the-personal-agent-protocol-is-not-a-standard-war-its-the-missing-identity-layer/</a></li>
<li>beri.net. &#8220;Meta&#8217;s Personal Agent Protocol Signs Walmart Before It Has a Spec.&#8221; <a href="https://www.beri.net/article/meta-sierra-personal-agent-protocol-oauth-guest-read-write-access-vs-ucp-acp-trusted-agent-protocol-retail-banks" target="_blank" rel="noopener nofollow">https://www.beri.net/article/meta-sierra-personal-agent-protocol-oauth-guest-read-write-access-vs-ucp-acp-trusted-agent-protocol-retail-banks</a></li>
<li>eesel.ai. &#8220;Personal Agent Protocol Explained: What Shipped, What&#8217;s Promised.&#8221; <a href="https://www.eesel.ai/blog/personal-agent-protocol" target="_blank" rel="noopener nofollow">https://www.eesel.ai/blog/personal-agent-protocol</a></li>
<li>Datafloq. &#8220;Meta and Sierra&#8217;s Personal Agent Protocol Starts With Login, Not Payments.&#8221; <a href="https://datafloq.com/meta-and-sierras-personal-agent-protocol-starts-with-login-not-payments/" target="_blank" rel="noopener nofollow">https://datafloq.com/meta-and-sierras-personal-agent-protocol-starts-with-login-not-payments/</a></li>
<li>DEV Community / TechAIWire. &#8220;Sierra and Meta Propose a Standard for Personal AI Agents.&#8221; <a href="https://dev.to/techaiwire/sierra-and-meta-propose-a-standard-for-personal-ai-agents-59p5" target="_blank" rel="noopener nofollow">https://dev.to/techaiwire/sierra-and-meta-propose-a-standard-for-personal-ai-agents-59p5</a></li>
<li>cellcog.ai. &#8220;Personal Agent Protocol: What Meta and Sierra Announced.&#8221; <a href="https://cellcog.ai/blog/personal-agent-protocol/" target="_blank" rel="noopener nofollow">https://cellcog.ai/blog/personal-agent-protocol/</a></li>
<li>The Next Web. &#8220;Sierra Announces Personal Agent Protocol.&#8221; <a href="https://thenextweb.com/news/personal-agent-protocol-sierra-meta" target="_blank" rel="noopener nofollow">https://thenextweb.com/news/personal-agent-protocol-sierra-meta</a></li>
<li>PPC Land. &#8220;Amazon Blocks Meta&#8217;s Muse AI Agent from Shopping on Its Site.&#8221; <a href="https://ppc.land/amazon-blocks-metas-muse-ai-agent-from-shopping-on-its-site/" target="_blank" rel="noopener nofollow">https://ppc.land/amazon-blocks-metas-muse-ai-agent-from-shopping-on-its-site/</a></li>
<li>eco.com. &#8220;Agent Payment Protocols Compared.&#8221; <a href="https://eco.com/support/en/articles/14839407-agent-payment-protocols-compared" target="_blank" rel="noopener nofollow">https://eco.com/support/en/articles/14839407-agent-payment-protocols-compared</a></li>
<li>AICoder. &#8220;Sierra and Meta Announce Personal Agent Protocol.&#8221; <a href="https://aicoder.com/news/news-20261007-sierra-meta-personal-agent-protocol-open-standard" target="_blank" rel="noopener nofollow">https://aicoder.com/news/news-20261007-sierra-meta-personal-agent-protocol-open-standard</a></li>
</ol>
]]></content:encoded>
					
					<wfw:commentRss>https://rpabotsworld.com/sierra-personal-agent-protocol-poppy-agentic-commerce-guide/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>Cisco Webex Agentic AI Platform: Dialog, Claude Managed Agents, and AI Agent 360 — The Complete Guide for Agentic AI Architects (2026)</title>
		<link>https://rpabotsworld.com/cisco-webex-agentic-ai-dialog-claude-agents-enterprise-guide/</link>
					<comments>https://rpabotsworld.com/cisco-webex-agentic-ai-dialog-claude-agents-enterprise-guide/#respond</comments>
		
		<dc:creator><![CDATA[Satish Prasad]]></dc:creator>
		<pubDate>Mon, 26 Oct 2026 05:08:57 +0000</pubDate>
				<category><![CDATA[AI Agents & Frameworks]]></category>
		<category><![CDATA[RPA & Bot Automation]]></category>
		<category><![CDATA[MCP]]></category>
		<guid isPermaLink="false">https://rpabotsworld.com/?p=32441</guid>

					<description><![CDATA[Cisco's Webex now hosts Claude Managed Agents, OpenAI dots via MCP, and Dialog for days-long customer workflows.]]></description>
										<content:encoded><![CDATA[<h1>Cisco Webex Agentic AI Platform: Dialog, Claude Managed Agents, and AI Agent 360 — The Complete Guide for Agentic AI Architects (2026)</h1>
<p>At WebexOne 2026 in Austin on October 7, Cisco did something no collaboration vendor has done before: it announced that AI agents — not copilots, not chatbots, but autonomous multi-step agents — can now be invited into Webex spaces, meetings, and calls as full participants. An analytics agent pulls usage stats mid-call. A presentation agent updates a PowerPoint deck while the team watches. A customer-service agent negotiates a refund internally for days after the customer hangs up, then sends an update on the customer&#8217;s preferred channel.</p>
<p>This is not incremental. Cisco is rebuilding Webex around three pillars — <strong>Claude Managed Agents</strong> (Anthropic), <strong>OpenAI dots</strong> via Model Context Protocol, and a new agentic harness called <strong>Dialog</strong> — while wrapping the entire stack in <strong>AI Agent 360</strong>, a governance and observability control plane built on Splunk. The result is the first enterprise collaboration platform where agents are first-class teammates, not sidebar assistants.</p>
<p>For RPA professionals and agentic AI architects, this matters for a concrete reason: the processes you automate today — ticket triage, data aggregation, report generation, customer follow-up — are exactly the workflows Cisco is targeting. The difference is that these agents live inside the communication layer where decisions happen, not in a separate orchestration tool that feeds results back into chat after the fact.</p>
<p>This guide breaks down every announced component, explains the architecture, maps the competitive landscape, and gives you a practical deployment framework — so you can assess whether Cisco&#8217;s agentic Webex belongs in your enterprise automation stack.</p>
<h2>What Cisco Announced at WebexOne 2026: The Full Picture</h2>
<p>Cisco&#8217;s announcements span collaboration software, customer experience infrastructure, hardware, and governance. Here is the complete lineup with confirmed timelines:</p>
<table>
<thead>
<tr>
<th>Component</th>
<th>Category</th>
<th>Status / Timeline</th>
</tr>
</thead>
<tbody>
<tr>
<td>Claude Managed Agents in Webex</td>
<td>Collaboration AI</td>
<td>Coming soon (early 2027)</td>
</tr>
<tr>
<td>OpenAI dots via MCP integration</td>
<td>Personal Agent</td>
<td>Future release</td>
</tr>
<tr>
<td>Dialog (agentic harness)</td>
<td>Customer Experience</td>
<td>Beta Q1 2027</td>
</tr>
<tr>
<td>AI Agent 360 + Splunk Observability</td>
<td>Governance</td>
<td>GA December 2026</td>
</tr>
<tr>
<td>Webex Contact Center MCP Server</td>
<td>Integration</td>
<td>GA October 2026</td>
</tr>
<tr>
<td>Agent-to-Agent protocol support</td>
<td>Interoperability</td>
<td>December 2026</td>
</tr>
<tr>
<td>NVIDIA Nemotron voice agents (BYO)</td>
<td>Voice AI</td>
<td>GA November 2026</td>
</tr>
<tr>
<td>Google Cloud Gemini Enterprise CX</td>
<td>Partner AI</td>
<td>Q4 2026</td>
</tr>
<tr>
<td>Amazon Nova 2.5 Sonic</td>
<td>Partner AI</td>
<td>Early access December 2026</td>
</tr>
<tr>
<td>SharePoint integration for AI Agent</td>
<td>Knowledge</td>
<td>GA November 2026</td>
</tr>
<tr>
<td>Universal Harness On-Premises</td>
<td>Data Sovereignty</td>
<td>Q1 2027</td>
</tr>
<tr>
<td>RoomOS 27</td>
<td>Devices</td>
<td>November 2026</td>
</tr>
<tr>
<td>Cisco Ceiling Speaker Pro</td>
<td>Hardware</td>
<td>Ordering Q4 2026</td>
</tr>
</tbody>
</table>
<p>The scope is deliberately broad. Cisco President and Chief Product Officer Jeetu Patel framed it this way: &#8220;The next era of work isn&#8217;t about bolting chatbots onto yesterday&#8217;s tools. It&#8217;s about AI agents working alongside people and teams.&#8221; That framing — agents as colleagues, not features — drives every architectural decision below.</p>
<h2>Claude Managed Agents in Webex: How Anthropic&#8217;s Agents Work Inside Collaboration</h2>
<p>The headline integration is <strong>Claude Managed Agents</strong>, Anthropic&#8217;s cloud-hosted agent infrastructure, embedded directly into Webex spaces, meetings, and calls. This is not a sidebar chatbot or a post-meeting summarizer. Claude agents participate in the flow of work as it happens.</p>
<h3>How It Works</h3>
<p>Teams interact with Claude agents through @mentions in Webex spaces — the same interaction model used for human colleagues. When mentioned, an agent can:</p>
<ul>
<li><strong>Analyze shared enterprise context</strong> — documents, conversation history, and data already in the space — without requiring users to copy-paste or re-explain</li>
<li><strong>Generate content</strong> — reports, summaries, code, proposals — using the team&#8217;s own context as grounding</li>
<li><strong>Coordinate multi-step workflows</strong> across Webex and connected third-party applications</li>
<li><strong>Produce recommendations</strong> and follow-up actions that other agents or humans can act on</li>
</ul>
<h3>The WebexOne Demo Workflow</h3>
<p>Cisco demonstrated a concrete workflow on stage: a user @mentions an analytics agent in a Webex space to pull usage statistics from a connected data source. The agent retrieves and formats the numbers. The user then @mentions a presentation agent, which takes those figures and updates a PowerPoint deck — all within the same space, during a live call. No tab-switching, no export-import, no manual formatting.</p>
<p>This is a significant step beyond what <a href="https://rpabotsworld.com/microsoft-copilot-autopilot-code-home-agentic-enterprise-guide/">Microsoft Copilot Autopilot</a> offers today. While Copilot operates as a single persistent agent within the Microsoft 365 ecosystem, Cisco&#8217;s model allows multiple specialized agents to collaborate within a single workspace — an analytics agent handing off to a presentation agent handing off to a scheduling agent — each doing what it does best.</p>
<h3>Why Anthropic, Not Just OpenAI?</h3>
<p>Cisco is not making Webex Claude-only. OpenAI&#8217;s dots product also gets MCP integration (discussed below). But Claude Managed Agents get the deeper, first-class integration for a reason: Claude&#8217;s architecture is optimized for enterprise context windows. With a 1M-token context window on Claude Haiku 5.5 (released October 7, 2026) and Claude&#8217;s strength in document analysis and structured reasoning, it maps well to the collaboration use case where agents need to reason over long meeting transcripts, multi-threaded conversations, and enterprise knowledge bases.</p>
<p>From Anthropic&#8217;s perspective, this is what one analyst called &#8220;a real distribution win&#8221; — getting Claude embedded in a platform used by hundreds of millions of enterprise workers, rather than requiring users to come to Claude&#8217;s own interface. For Cisco, it hedges model risk: enterprises generally avoid single-vendor AI lock-in, and offering both Claude and dots signals openness.</p>
<h2>OpenAI dots in Webex: MCP-Powered Personal Agents</h2>
<p>The second agent integration uses <strong>Model Context Protocol (MCP)</strong> to bring OpenAI&#8217;s always-on dots agents into Webex. If you&#8217;ve been tracking <a href="https://rpabotsworld.com/openai-devday-2026-dots-agents-gpt-6-1-sol-guide/">OpenAI&#8217;s DevDay 2026 announcements</a>, you know dots are persistent agents built on GPT-6 Astra that continue working between conversations, with their own cloud compute and memory.</p>
<h3>How the MCP Integration Works</h3>
<p>The <strong>Webex Suite MCP integration</strong> exposes Webex context — spaces, messages, calendar events, call logs — to dots agents through the <a href="https://rpabotsworld.com/a2a-joins-mcp-aaif-agentic-ai-architects-guide/">Model Context Protocol</a>. This means a dots agent can:</p>
<ul>
<li>Read and reason over Webex conversations and files</li>
<li>Pull context from meetings and calendar events</li>
<li>Take actions within Webex based on its ongoing goals</li>
</ul>
<p>Cisco plans to go further: a future release will make dots available <em>inside</em> Webex as an always-on personal agent. Think of it as a persistent digital colleague that answers your calls when you&#8217;re busy, assesses urgency, schedules follow-ups, and sends messages on your behalf — all drawing on the full context of your Webex workspace.</p>
<h3>Why MCP Matters Here</h3>
<p>For agentic AI architects, the MCP integration is architecturally significant. It means Cisco is not building a walled garden. Any agent platform that speaks MCP can, in principle, interact with Webex context. The <strong>Webex Contact Center MCP Server</strong> — set for GA in October 2026 — makes this concrete: it&#8217;s an MCP endpoint that exposes contact center data and actions to compliant agents, whether from OpenAI, Anthropic, or an open-source framework like LangChain or CrewAI.</p>
<h2>Dialog: The Agentic Harness That Changes Customer Experience</h2>
<p>If Claude Managed Agents transform internal collaboration, <strong>Dialog</strong> transforms customer-facing operations. Dialog is what Cisco calls an &#8220;agentic harness&#8221; — a software layer beneath the Webex AI Agent platform that turns AI agents into persistent relationship managers, not just conversation handlers.</p>
<h3>The Four Pillars of Dialog</h3>
<p>Dialog&#8217;s architecture rests on four capabilities that, taken together, represent a fundamentally different approach to customer service automation:</p>
<p><strong>1. Relationship Context (the &#8220;Relationship Cortex&#8221;)</strong></p>
<p>Dialog merges enterprise data, customer history, and Webex CX data into what Cisco calls a unified &#8220;relationship cortex.&#8221; This is not a CRM lookup. It is a reasoning layer that understands customer lifetime value, interaction patterns, preferences, and history across every channel — phone, messaging, email, in-person. When an agent picks up a case, it already knows who the customer is, what they&#8217;ve tried before, and what matters to them.</p>
<p><strong>2. Agent Onboarding</strong></p>
<p>Rather than building agents from hand-crafted conversation flows (the traditional IVR/chatbot approach that RPA teams know well), Dialog agents are onboarded by indexing existing employee handbooks, knowledge bases, operating procedures, and policies. Cisco describes this as treating AI agents &#8220;like new teammates&#8221; — you give them the documentation, they learn the rules, exceptions, and escalation paths.</p>
<p>For <a href="https://rpabotsworld.com/automation-anywhere-vs-uipath-agentic-procure-to-pay-2026/">RPA teams migrating from rule-based automation</a>, this is the shift in practice: instead of mapping every decision branch in a process designer, you point the agent at the documentation and let it reason about edge cases.</p>
<p><strong>3. Self-Learning Loop</strong></p>
<p>Every interaction teaches Dialog something. The system learns from customer outcomes and employee judgment, within enterprise policy limits. This is not unsupervised drift — the learning runs within guardrails set by Cisco AI Defense — but it means agents get better at handling cases over time, much like a human agent who gets faster at recognizing patterns after hundreds of similar calls.</p>
<p><strong>4. Long-Horizon Task Execution</strong></p>
<p>This is the capability that most clearly separates Dialog from traditional contact center automation. A Dialog agent doesn&#8217;t just handle the conversation — it keeps working on the customer&#8217;s behalf for days after the call ends.</p>
<h3>The WebexOne Demo: What Long-Horizon Execution Looks Like</h3>
<p>Cisco&#8217;s on-stage demo told a specific story: a customer calls about a missed flight and asks for a refund that standard policy doesn&#8217;t cover. Here is what the Dialog agent does:</p>
<ol>
<li><strong>Gathers context</strong> — the customer&#8217;s status, lifetime value, recent travel disruptions, and the specific policy exception process</li>
<li><strong>Tells the customer</strong> they no longer need to chase the issue — the agent will handle it</li>
<li><strong>Opens an internal ticket</strong> and begins negotiating with a human escalation specialist, citing the customer&#8217;s value and recent service failures as grounds for an exception</li>
<li><strong>Checks insurance details</strong> by contacting the customer&#8217;s own personal AI agent (agent-to-agent communication)</li>
<li><strong>Runs for days</strong>, coordinating across multiple internal rounds with staff, other agents, and back-office systems</li>
<li><strong>Sends daily updates</strong> to the customer on their preferred channel</li>
</ol>
<p>As VP and GM of Webex Customer Experience Vinod Muthukrishnan put it: &#8220;It&#8217;s demonstrating agency, but it&#8217;s not making stuff up.&#8221; The agent operates within policy boundaries, but it pursues the customer&#8217;s outcome proactively — a fundamentally different model from reactive ticket queues.</p>
<p>This vision — an agent that acts as a customer advocate across time and systems — is what Muthukrishnan calls &#8220;managing relationships: thousands of conversations, hundreds of journeys, but one relationship.&#8221; If you&#8217;re familiar with <a href="https://rpabotsworld.com/salesforce-agentforce-job-ready-agents-long-horizon-runtime-guide-3/">Salesforce&#8217;s Agentforce long-horizon runtime</a>, Dialog is Cisco&#8217;s answer: a persistent execution engine for customer-facing workflows that doesn&#8217;t stop when the channel closes.</p>
<h2>AI Agent 360 and Splunk Observability: Enterprise Governance for Agents</h2>
<p>Every agent platform needs a governance story, and Cisco&#8217;s is more concrete than most — because Cisco owns Splunk.</p>
<h3>What AI Agent 360 Does</h3>
<p><strong>AI Agent 360</strong> is the control plane for the entire Webex AI Agent platform. It covers the full lifecycle:</p>
<ul>
<li><strong>Design and testing:</strong> Build, connect, and test agents in AI Agent Studio before deployment</li>
<li><strong>Deployment:</strong> Roll out agents with defined permissions, data access boundaries, and escalation rules</li>
<li><strong>Monitoring:</strong> Real-time telemetry on agent behavior, decision traces, and conversation outcomes</li>
<li><strong>Optimization:</strong> Continuous evaluation and red-teaming from a single interface</li>
</ul>
<h3>Agent Observability with Splunk</h3>
<p>The December 2026 GA of <strong>Agent Observability with Splunk</strong> adds a layer that most competing platforms lack: full-stack observability that links agent behavior to business outcomes. This is not just &#8220;did the agent respond?&#8221; — it is:</p>
<ul>
<li><strong>Agent health:</strong> Is the agent performing as expected? Where is it struggling?</li>
<li><strong>Escalation patterns:</strong> When and why do agents hand off to humans? What does that cost?</li>
<li><strong>Business impact:</strong> What is the resolution rate? How does agent performance correlate with customer satisfaction and revenue?</li>
</ul>
<p>Muthukrishnan frames governance as the decisive factor: &#8220;It&#8217;s never been harder to make it enterprise grade.&#8221; And the threat surface is real — he estimates the opportunity surface grew 10x with agentic AI, but the threat surface grew 100x. The Splunk integration is Cisco&#8217;s answer: if you can observe it, you can govern it.</p>
<p>For automation architects building agent governance frameworks, this is worth studying alongside <a href="https://rpabotsworld.com/nvidia-open-agent-safety-platform-openshell-sentry-guide/">NVIDIA&#8217;s Open Agent Safety Platform</a>. Where NVIDIA focuses on safety guardrails at the model layer, Cisco focuses on operational governance at the platform layer — they are complementary, not competing.</p>
<h3>Cisco AI Defense</h3>
<p>Sitting alongside AI Agent 360 is <strong>Cisco AI Defense</strong>, which provides security guardrails for agent behavior. Teams can create custom guardrails using natural-language instructions — &#8220;never share customer payment details with third-party agents,&#8221; for example — rather than writing complex policy rules. AI Defense determines what agents can access, what they can do, and how they act, with identity, security, and visibility controls applied uniformly across all agent types.</p>
<h2>The Multi-Model, Multi-Vendor Architecture: Why Cisco Is Hedging</h2>
<p>One of the most architecturally interesting aspects of Cisco&#8217;s strategy is its deliberate multi-vendor approach. The Webex AI Agent platform does not bet on a single model provider:</p>
<table>
<thead>
<tr>
<th>Provider</th>
<th>Integration</th>
<th>Use Case</th>
</tr>
</thead>
<tbody>
<tr>
<td>Anthropic (Claude)</td>
<td>Claude Managed Agents</td>
<td>Enterprise collaboration, analysis, content generation</td>
</tr>
<tr>
<td>OpenAI (dots)</td>
<td>MCP integration</td>
<td>Personal always-on agents</td>
</tr>
<tr>
<td>NVIDIA</td>
<td>Nemotron + Parakeet</td>
<td>Voice AI, multilingual agents</td>
</tr>
<tr>
<td>Google Cloud</td>
<td>Gemini Enterprise</td>
<td>CX agents</td>
</tr>
<tr>
<td>Amazon</td>
<td>Nova 2.5 Sonic</td>
<td>Voice agents (beta)</td>
</tr>
<tr>
<td>On-premises (BYO)</td>
<td>Universal Harness</td>
<td>Data sovereignty, compliance</td>
</tr>
</tbody>
</table>
<p>This is not accidental. Enterprise IT departments demand model portability, and Cisco&#8217;s collaboration hardware installed base — the devices in conference rooms worldwide — gives it a distribution channel that pure-software competitors cannot match. Aragon Research analysts Jim Lundy and Adam Pease noted that &#8220;Cisco&#8217;s endpoint hardware footprint is a differentiator that rivals without comparable hardware will struggle to match.&#8221;</p>
<p>The <strong>Universal Harness On-Premises</strong> option, coming in Q1 2027, is particularly relevant for regulated industries. It lets organizations host third-party AI models locally for data sovereignty and compliance — a capability that <a href="https://rpabotsworld.com/ibm-watsonx-orchestrate-agentic-control-plane-guide/">IBM watsonx Orchestrate</a> also emphasizes, but that most SaaS-first agent platforms cannot offer.</p>
<h2>What This Means for RPA Teams: The Collaboration-Layer Shift</h2>
<p>If you build RPA workflows today, here is the concrete impact:</p>
<h3>1. Processes Move Into the Communication Layer</h3>
<p>Traditional RPA lives in the back office — attended bots on desktops, unattended bots in server rooms, processing data between systems. Cisco&#8217;s agentic Webex moves process automation into the place where decisions are made: the meeting room, the team chat, the customer call. The analytics-to-PowerPoint demo is trivial in isolation, but the pattern it establishes — agents executing multi-step work inside the collaboration flow — is a direct challenge to the &#8220;RPA processes data, humans communicate results&#8221; model.</p>
<h3>2. Contact Center Automation Gets a Long Memory</h3>
<p>Dialog&#8217;s long-horizon execution model is the biggest shift for RPA teams doing contact center automation. Today, most contact center bots handle a single interaction and then forget. Dialog agents maintain persistent context and pursue outcomes across days, channels, and internal systems. If your organization runs customer-service RPA — <a href="https://rpabotsworld.com/forrester-wave-adaptive-process-orchestration-q3-2026-guide/">particularly in process orchestration</a> — Dialog represents a fundamentally different approach that may reduce the need for separate bot-to-human handoff workflows.</p>
<h3>3. MCP Becomes the Integration Standard</h3>
<p>Cisco&#8217;s adoption of MCP for the Webex Contact Center Server and the dots integration adds another major enterprise vendor to the MCP ecosystem. For RPA architects already building MCP-compliant tool integrations, this means your agents can potentially interact with Webex context — spaces, meetings, contact center data — through a standard protocol rather than a proprietary API.</p>
<h3>4. Governance Moves from Afterthought to Architecture</h3>
<p>AI Agent 360 with Splunk observability sets a new bar for what enterprise buyers will expect from any agent platform. If you are building agent solutions on UiPath, Automation Anywhere, or open-source frameworks, expect your enterprise customers to ask: &#8220;How do we observe agent behavior end-to-end? How do we link agent actions to business outcomes?&#8221; Cisco just made those questions standard procurement criteria.</p>
<h2>Competitive Landscape: How Cisco Stacks Up</h2>
<p>Cisco is not entering the agentic AI market alone. Here is how its approach compares to the major players:</p>
<table>
<thead>
<tr>
<th>Dimension</th>
<th>Cisco Webex</th>
<th>Microsoft (Copilot/Teams)</th>
<th>Salesforce (Agentforce)</th>
<th>Google (Gemini Agents)</th>
</tr>
</thead>
<tbody>
<tr>
<td>Agent model</td>
<td>Multi-vendor (Claude, dots, NVIDIA, Gemini, Nova)</td>
<td>Single-vendor (GPT family)</td>
<td>Multi-model via Einstein</td>
<td>Gemini-first, with partner support</td>
</tr>
<tr>
<td>Primary surface</td>
<td>Collaboration (meetings, spaces, calls) + CX</td>
<td>Productivity (Office 365, Teams)</td>
<td>CRM and customer operations</td>
<td>Workspace (Docs, Sheets, Meet)</td>
</tr>
<tr>
<td>Long-horizon execution</td>
<td>Dialog (days-long tasks, Q1 2027 beta)</td>
<td>Copilot Autopilot (persistent agent)</td>
<td>Long-Horizon Runtime (GA)</td>
<td>Gemini Enterprise Agent Platform</td>
</tr>
<tr>
<td>Observability</td>
<td>Splunk-native (owned)</td>
<td>Azure Monitor / third-party</td>
<td>Einstein Trust Layer</td>
<td>Google Cloud Monitoring</td>
</tr>
<tr>
<td>On-premises option</td>
<td>Universal Harness (Q1 2027)</td>
<td>Azure Arc / hybrid</td>
<td>Limited (Hyperforce)</td>
<td>GDC / Anthos</td>
</tr>
<tr>
<td>Hardware integration</td>
<td>RoomOS 27, Ceiling Speaker Pro, Cisco devices</td>
<td>Surface Hub, Teams Rooms</td>
<td>None</td>
<td>Google Meet hardware</td>
</tr>
<tr>
<td>MCP support</td>
<td>Contact Center MCP Server, dots MCP</td>
<td>Limited</td>
<td>Via ecosystem</td>
<td>A2A protocol (related)</td>
</tr>
</tbody>
</table>
<p>Cisco&#8217;s differentiators are its <strong>multi-model openness</strong>, its <strong>owned observability stack</strong> (Splunk), and its <strong>physical hardware reach</strong>. Its risks are execution timelines — many of the most interesting capabilities (Claude Managed Agents, Dialog, Universal Harness) are &#8220;coming soon&#8221; or in beta, not GA — and the reliability gap that the Engadget article flagged: multi-step agent workflows currently fail 60–70% of the time across the industry, and a 95% per-step success rate still yields only 36% end-to-end success over a 20-step workflow.</p>
<h2>Practical Deployment Framework: How to Evaluate Cisco&#8217;s Agentic Webex for Your Organization</h2>
<p>Based on Cisco&#8217;s own recommendations, analyst guidance from Aragon Research, and the practical realities of enterprise agent deployment, here is a structured approach:</p>
<h3>Step 1: Identify Your High-Value Collaboration Workflows</h3>
<p>Start with workflows where decisions happen inside Webex today — not in a separate system. Examples: weekly business reviews that require pulling data from multiple sources, customer escalation processes that span multiple channels, project status updates that involve compiling inputs from several teams. These are the workflows where in-meeting agents deliver immediate value.</p>
<h3>Step 2: Assess Your Data Governance Readiness</h3>
<p>As Aragon Research recommends: strengthen data governance, security controls, and identity management <em>before</em> deploying agents. Connecting agents to multiple applications makes data permissions and oversight critical. Define what agents can access, what they can change, and what requires human approval.</p>
<h3>Step 3: Prepare Your Knowledge Bases</h3>
<p>Dialog&#8217;s onboarding model works by indexing existing documentation. If your knowledge bases are outdated, incomplete, or contradictory, agents will inherit those problems. Audit and clean your handbooks, SOPs, and policy documents before agent onboarding. This is the same &#8220;garbage in, garbage out&#8221; principle that RPA teams know from process mining — except agents can propagate bad documentation at scale.</p>
<h3>Step 4: Start With a Bounded Pilot</h3>
<p>TechRepublic recommends choosing a single workflow to test, confirming feature availability and licensing, and starting with a limited pilot that requires human approval for all sensitive actions. Do not attempt a broad rollout with agents that are still in beta (Dialog) or &#8220;coming soon&#8221; (Claude Managed Agents).</p>
<h3>Step 5: Build Your Observability Baseline</h3>
<p>When AI Agent 360 with Splunk observability reaches GA in December 2026, establish baseline metrics before expanding: agent resolution rate, escalation frequency, average handling time, and customer satisfaction correlation. You cannot optimize what you do not measure.</p>
<h2>What to Watch: Timeline and Risks</h2>
<p>The biggest risk in Cisco&#8217;s announcement is the gap between vision and availability:</p>
<ul>
<li><strong>Available now or soon (Q4 2026):</strong> Webex Contact Center MCP Server, AI Agent for UCCE, RoomOS 27, AI Agent 360 with Splunk, Agent-to-Agent protocol, NVIDIA Nemotron voice agents, SharePoint integration</li>
<li><strong>Coming in Q1 2027:</strong> Claude Managed Agents (full capabilities), Dialog beta, Universal Harness On-Premises, deeper dots integration</li>
<li><strong>Timeline unclear:</strong> Dots as always-on personal agent inside Webex, full Google Cloud Gemini CX integration</li>
</ul>
<p>Enterprise buyers should not plan production workloads around Q1 2027 betas. But they <em>should</em> be preparing: cleaning knowledge bases, defining agent governance policies, and running pilots on the capabilities that are available now (the Contact Center MCP Server and AI Agent for UCCE) so they are ready to move quickly when Dialog and Claude Managed Agents reach GA.</p>
<h2>Frequently Asked Questions</h2>
<h3>What is Cisco Dialog and how does it differ from traditional contact center chatbots?</h3>
<p>Dialog is an agentic harness — a software layer beneath Cisco&#8217;s Webex AI Agent platform — that enables long-horizon task execution for customer experience. Unlike traditional chatbots that handle a single conversation and forget, Dialog agents maintain persistent customer context across channels and time, continue working on the customer&#8217;s behalf after a conversation ends (for days if needed), and coordinate across people, other agents, and backend systems to resolve complex issues. Dialog learns from each interaction within enterprise policy limits and onboards by indexing existing documentation rather than requiring hand-built conversation flows.</p>
<h3>Does Cisco Webex only use Claude, or can I bring my own AI model?</h3>
<p>Cisco&#8217;s agentic Webex platform is deliberately multi-model. It integrates Claude Managed Agents (Anthropic), OpenAI dots via MCP, NVIDIA Nemotron and Parakeet for voice AI, Google Cloud Gemini for CX agents, and Amazon Nova Sonic. The Universal Harness On-Premises option (Q1 2027) will also let organizations host their own third-party models locally for data sovereignty. This multi-vendor approach avoids model lock-in and lets enterprises choose the best model for each use case.</p>
<h3>When will Claude Managed Agents be available in Webex?</h3>
<p>Cisco has announced Claude Managed Agents as &#8220;coming soon&#8221; with full capabilities expected in early 2027. The Webex Contact Center MCP Server and AI Agent for Unified Contact Center Enterprise are available now. AI Agent 360 with Splunk observability reaches GA in December 2026. Dialog enters beta in Q1 2027.</p>
<h3>How does Cisco handle AI agent security and governance in Webex?</h3>
<p>Cisco uses a two-layer governance approach: Cisco AI Defense provides security guardrails using natural-language policy instructions (e.g., &#8220;never share payment details with third-party agents&#8221;), while AI Agent 360 with Splunk observability provides real-time telemetry, decision traces, escalation pattern analysis, and business impact measurement. Teams can evaluate, red-team, and monitor agents from a single control plane, with identity and access controls applied uniformly across all agent types.</p>
<h3>Can Cisco Webex agentic AI replace my existing RPA workflows?</h3>
<p>Not directly, but it will reshape how some processes work. Cisco&#8217;s agentic Webex is strongest where decisions happen inside collaboration tools — meetings, team chats, customer calls. Processes that involve pulling data during discussions, updating reports in real time, or managing long-running customer cases are natural fits. Back-office processes that run without human interaction (batch data processing, system-to-system transfers) will still run better on traditional RPA or process orchestration platforms. The more likely outcome is coexistence: agentic Webex for the collaboration and CX layers, RPA for structured back-office automation, with MCP as the integration bridge.</p>
<h2>Key Takeaways</h2>
<ul>
<li><strong>Cisco is making agents first-class Webex participants</strong> — able to join spaces, meetings, and calls, not just summarize them after the fact</li>
<li><strong>Claude Managed Agents and OpenAI dots</strong> handle internal collaboration; <strong>Dialog</strong> handles customer-facing workflows — together they cover both sides of enterprise communication</li>
<li><strong>Dialog&#8217;s long-horizon execution</strong> lets agents work on customer issues for days, across channels and internal systems, fundamentally changing contact center automation</li>
<li><strong>AI Agent 360 with Splunk</strong> sets a new bar for agent observability — linking agent behavior to escalation patterns and business outcomes</li>
<li><strong>The multi-model architecture</strong> (Claude, dots, NVIDIA, Gemini, Nova, plus BYO on-premises) avoids vendor lock-in and gives enterprises model choice per use case</li>
<li><strong>MCP adoption</strong> (Contact Center MCP Server, dots integration) signals Cisco is building for interoperability, not a walled garden</li>
<li><strong>Most headline features ship Q4 2026 through Q1 2027</strong> — plan pilots on what&#8217;s GA now and prepare governance, knowledge bases, and observability for what&#8217;s coming</li>
</ul>
<h2>External References</h2>
<ol>
<li>Cisco Newsroom, &#8220;Cisco Unveils New Agentic Collaboration Experiences,&#8221; October 7, 2026 — <a href="https://newsroom.cisco.com/c/r/newsroom/en/us/a/y2026/m10/cisco-unveils-new-agentic-collaboration-experiences.html" target="_blank" rel="noopener nofollow">newsroom.cisco.com</a></li>
<li>TechRepublic, &#8220;Cisco Announces Webex AI Agents: What IT Should Know,&#8221; October 2026 — <a href="https://www.techrepublic.com/article/news-cisco-webex-ai-agents-meetings-customer-service/" target="_blank" rel="noopener nofollow">techrepublic.com</a></li>
<li>Engadget, &#8220;Cisco Is Bringing a New Agentic Framework to Its Webex Enterprise Platform,&#8221; October 2026 — <a href="https://www.engadget.com/2278872/cisco-is-bringing-a-new-agentic-framework-to-its-webex-enterprise-platform/" target="_blank" rel="noopener nofollow">engadget.com</a></li>
<li>SiliconANGLE, &#8220;Cisco Broadens Agentic Features in Webex Collaboration Platform,&#8221; October 7, 2026 — <a href="https://siliconangle.com/2026/10/07/cisco-broadens-agentic-features-in-webex-collaboration-platform/" target="_blank" rel="noopener nofollow">siliconangle.com</a></li>
<li>Aragon Research (Jim Lundy, Adam Pease), &#8220;Cisco Webex Strategy: Driven by Agentic AI,&#8221; October 8, 2026 — <a href="https://aragonresearch.com/cisco-webex-strategy-driven-by-agentic-ai/" target="_blank" rel="noopener nofollow">aragonresearch.com</a></li>
<li>Startup Fortune, &#8220;Cisco Turns Webex Into an AI Agent Workspace With Claude Built In,&#8221; October 2026 — <a href="https://startupfortune.com/cisco-turns-webex-into-an-ai-agent-workspace-with-claude-built-in/" target="_blank" rel="noopener nofollow">startupfortune.com</a></li>
<li>CX Today, &#8220;WebexOne 2026: Cisco Dialog Turns AI Agents Into Advocates,&#8221; October 2026 — <a href="https://cxtoday.com/webexone-2026-cisco-dialog-turns-ai-agents-into-advocates" target="_blank" rel="noopener nofollow">cxtoday.com</a></li>
<li>Unite.AI, &#8220;Cisco Debuts New-Look Webex AI Agent Platform,&#8221; October 2026 — <a href="https://www.unite.ai/?p=481220" target="_blank" rel="noopener nofollow">unite.ai</a></li>
<li>CX Foundation, &#8220;Cisco Debuts New-Look Webex AI Agent Platform,&#8221; October 2026 — <a href="https://cxfoundation.com/news/cisco-debuts-new-look-webex-ai-agent-platform" target="_blank" rel="noopener nofollow">cxfoundation.com</a></li>
</ol>
]]></content:encoded>
					
					<wfw:commentRss>https://rpabotsworld.com/cisco-webex-agentic-ai-dialog-claude-agents-enterprise-guide/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>Microsoft Work IQ and the Work IQ MCP Server: The Complete Guide for Agentic AI Architects (2026)</title>
		<link>https://rpabotsworld.com/microsoft-work-iq-mcp-server-dynamics-365-agentic-ai-guide/</link>
					<comments>https://rpabotsworld.com/microsoft-work-iq-mcp-server-dynamics-365-agentic-ai-guide/#respond</comments>
		
		<dc:creator><![CDATA[Satish Prasad]]></dc:creator>
		<pubDate>Mon, 26 Oct 2026 05:08:51 +0000</pubDate>
				<category><![CDATA[AI Agents & Frameworks]]></category>
		<category><![CDATA[MCP]]></category>
		<guid isPermaLink="false">https://rpabotsworld.com/?p=32452</guid>

					<description><![CDATA[Microsoft Work IQ and the Work IQ MCP Server: The Complete Guide for Agentic AI Architects (2026) An enterprise sales manager asks Copilot a straightforward question: &#8220;Which of my renewals are at risk?&#8221; The CRM shows green across the board — every deal is on track. But behind that clean dashboard, a service ticket has [&#8230;]]]></description>
										<content:encoded><![CDATA[<h1>Microsoft Work IQ and the Work IQ MCP Server: The Complete Guide for Agentic AI Architects (2026)</h1>
<p>An enterprise sales manager asks Copilot a straightforward question: &#8220;Which of my renewals are at risk?&#8221; The CRM shows green across the board — every deal is on track. But behind that clean dashboard, a service ticket has been open for three weeks, the customer&#8217;s last two emails carried a frustrated tone, and a custom equipment app logged a failed deployment. The CRM never saw any of that. Copilot, grounded only in the CRM, gives a confidently wrong answer.</p>
<p>This is the context gap that kills enterprise agent projects before they deliver value. According to Microsoft, it&#8217;s also the exact problem that <strong>Work IQ</strong> — the intelligence layer beneath Copilot and AI agents — was built to solve. With its public preview rolling out through October 2026 and ERP support landing in late October, Work IQ represents Microsoft&#8217;s most consequential bet on what separates toy agents from production agents: not better models, but better business context.</p>
<p>This guide breaks down the architecture, the MCP server, the governance model, and the practical implications for RPA and automation teams evaluating whether to build their next generation of agents on the Microsoft stack.</p>
<h2>What Is Microsoft Work IQ?</h2>
<p>Work IQ is the shared business-context layer that grounds Microsoft Copilot and AI agents in an organization&#8217;s actual data — email, calendar, Teams conversations, Dynamics 365 records, Power Platform apps, files, and people. Rather than forcing each agent to build its own retrieval pipeline, Work IQ provides a single semantic model that every agent in the organization can share.</p>
<p>It sits within the broader <strong>Microsoft IQ</strong> umbrella, announced at Build 2026 on June 2. That umbrella includes four engines that compose together:</p>
<table>
<thead>
<tr>
<th>Engine</th>
<th>What It Does</th>
<th>Data Sources</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Work IQ</strong></td>
<td>Organizational context — how people work, who they work with, what they&#8217;re working on</td>
<td>Email, calendar, meetings, Teams chats, files, Dynamics 365, Power Apps</td>
</tr>
<tr>
<td><strong>Foundry IQ</strong></td>
<td>Institutional knowledge retrieval with SLA-backed endpoints</td>
<td>Knowledge bases, policy documents, internal wikis, Azure SQL, MCP sources</td>
</tr>
<tr>
<td><strong>Fabric IQ</strong></td>
<td>Live operational data with permission inheritance</td>
<td>OneLake, Power BI datasets, semantic models</td>
</tr>
<tr>
<td><strong>Web IQ</strong></td>
<td>Real-time external web grounding (sub-165ms latency, zero data retention)</td>
<td>Public internet</td>
</tr>
</tbody>
</table>
<p>Work IQ reached general availability on June 16, 2026, with its core APIs, A2A (agent-to-agent) protocol support, and redesigned remote MCP server. The October 2026 milestone extends its grounding to <strong>Dynamics 365 Sales and Customer Service, Power Apps, and third-party line-of-business applications</strong> — and late October adds Dynamics 365 Finance and Operations, bringing ERP data into the same context layer.</p>
<p>For teams that have been building <a href="https://rpabotsworld.com/a2a-joins-mcp-aaif-agentic-ai-architects-guide/">agents around the A2A and MCP protocols</a>, Work IQ is significant because it&#8217;s not just another API — it&#8217;s Microsoft&#8217;s answer to the question of how agents understand a business well enough to take action on its behalf.</p>
<h2>The Architecture: How Work IQ Grounds Agents in Business Data</h2>
<p>Work IQ&#8217;s architecture has four layers, each solving a distinct piece of the &#8220;make agents actually useful in enterprise&#8221; problem:</p>
<h3>1. The Chat Layer</h3>
<p>The top layer supports two integration patterns: <strong>A2A</strong> (agent-to-agent protocol) for agent delegation and <strong>REST</strong> for web application interactions. Responses are fully processed and carry the same quality as Copilot&#8217;s own answers. This means a third-party agent built on LangChain or CrewAI can call Work IQ through A2A and get back contextually grounded responses — not raw data that the agent has to interpret from scratch.</p>
<h3>2. The Context Layer</h3>
<p>This is where Work IQ does its heaviest lifting. It assembles and grounds context internally, so agents never need to build their own retrieval pipelines. The context layer draws from:</p>
<ul>
<li><strong>A semantic model</strong> derived from existing Dynamics 365 and Power Apps artifacts — views, forms, entity relationships, and a business glossary for company-specific terminology</li>
<li><strong>Microsoft 365 signals</strong> — email threads, Teams conversations, calendar events, shared files, and collaboration patterns</li>
<li><strong>Third-party LOB applications</strong> connected through the platform</li>
</ul>
<p>Critically, every agent using the semantic model shares the same interpretation of entities like accounts, contracts, and service cases. This eliminates the &#8220;agent A thinks &#8216;customer&#8217; means a CRM contact while agent B thinks it means an Azure AD identity&#8221; problem that plagues multi-agent deployments.</p>
<h3>3. The Tools Layer</h3>
<p>A compact set of generic tools and business skills for reasoning, retrieval, and actions across Microsoft 365. More on this in the MCP server section below.</p>
<h3>4. The Workspaces Layer</h3>
<p>SharePoint Embedded working storage within the Microsoft 365 tenant boundary. Agents use these workspaces for intermediate data, outputs, and handoffs between agents. This is governed by existing information protection policies — Purview sensitivity labels, Conditional Access, and DLP rules all apply to agent working storage the same way they apply to human document libraries.</p>
<h2>The Work IQ MCP Server: 12 Tools Instead of Hundreds</h2>
<p>The Work IQ MCP server is where the architecture meets the developer experience — and it&#8217;s the component that matters most to automation architects deciding between building agents on Microsoft&#8217;s stack versus alternatives like <a href="https://rpabotsworld.com/salesforce-agentforce-job-ready-agents-long-horizon-runtime-guide-3/">Salesforce Agentforce</a> or <a href="https://rpabotsworld.com/google-gemini-enterprise-agent-platform-guide/">Google&#8217;s Gemini Enterprise Agent Platform</a>.</p>
<p>The core design principle: <strong>fewer tools, more paths</strong>. Instead of exposing hundreds of granular Microsoft Graph operations as individual tools (a design that overwhelms LLMs during tool selection and burns tokens on schema loading), Work IQ consolidates everything into 12 generic tools organized in three categories:</p>
<h3>Entity Tools (7)</h3>
<table>
<thead>
<tr>
<th>Tool</th>
<th>Purpose</th>
<th>Example</th>
</tr>
</thead>
<tbody>
<tr>
<td><code>fetch</code></td>
<td>Read any M365 resource</td>
<td><code>fetch /me/messages</code> → read emails</td>
</tr>
<tr>
<td><code>fetch_blob</code></td>
<td>Retrieve binary content (files, attachments)</td>
<td>Download a SharePoint document</td>
</tr>
<tr>
<td><code>create_entity</code></td>
<td>Create a new resource</td>
<td><code>create_entity /me/events</code> → create calendar event</td>
</tr>
<tr>
<td><code>update_entity</code></td>
<td>Modify an existing resource</td>
<td>Update a Dynamics 365 opportunity stage</td>
</tr>
<tr>
<td><code>delete_entity</code></td>
<td>Remove a resource</td>
<td>Delete a draft email</td>
</tr>
<tr>
<td><code>do_action</code></td>
<td>Trigger an operation</td>
<td><code>do_action /me/sendMail</code> → send an email</td>
</tr>
<tr>
<td><code>call_function</code></td>
<td>Invoke a function endpoint</td>
<td><code>call_function /search/query</code> → semantic search</td>
</tr>
</tbody>
</table>
<h3>Copilot Tools (3)</h3>
<table>
<thead>
<tr>
<th>Tool</th>
<th>Purpose</th>
</tr>
</thead>
<tbody>
<tr>
<td><code>ask</code></td>
<td>Invoke Microsoft 365 Copilot — &#8220;What deals closed this quarter?&#8221;</td>
</tr>
<tr>
<td><code>retrieve</code></td>
<td>Return grounding content and source metadata (public preview)</td>
</tr>
<tr>
<td><code>list_agents</code></td>
<td>Discover available agents in the organization</td>
</tr>
</tbody>
</table>
<h3>Schema Tools (2)</h3>
<table>
<thead>
<tr>
<th>Tool</th>
<th>Purpose</th>
</tr>
</thead>
<tbody>
<tr>
<td><code>get_schema</code></td>
<td>Retrieve OpenAPI schemas at runtime — agents discover data structures on demand instead of loading them into context</td>
</tr>
<tr>
<td><code>search_paths</code></td>
<td>Discover available API paths — agents can find what&#8217;s accessible without hardcoding endpoints</td>
</tr>
</tbody>
</table>
<p>The path-based model means that <strong>adding new workloads to Work IQ adds paths, not tools</strong>. When Dynamics 365 Finance and Operations joins the preview in late October, agents don&#8217;t need new tools — they get new resource paths accessible through the same <code>fetch</code>, <code>create_entity</code>, and <code>update_entity</code> verbs they already know.</p>
<p>All requests flow through a single MCP endpoint: <code>https://workiq.svc.cloud.microsoft/mcp</code>.</p>
<h3>Developer Setup</h3>
<p>The CLI and VS Code integration make setup lightweight:</p>
<pre><code># Install the CLI
npm install -g @microsoft/workiq

# Or run directly with npx
npx -y @microsoft/workiq mcp

# Accept the EULA (required before first use)
workiq accept-eula</code></pre>
<p>VS Code MCP configuration:</p>
<pre><code>{
  "workiq": {
    "command": "npx",
    "args": ["-y", "@microsoft/workiq", "mcp"],
    "tools": ["*"]
  }
}</code></pre>
<p>Authentication uses Microsoft Entra ID with delegated permissions. MCP clients discover the auth configuration automatically via <code>/.well-known/oauth-protected-resource</code>. One important constraint: <strong>application-only authentication is not supported</strong> — every request runs as a signed-in user, which is a deliberate governance choice rather than a technical limitation.</p>
<h2>Business Skills: Codified Playbooks for Agents</h2>
<p>If the MCP server gives agents access to data, <strong>business skills</strong> give agents access to process knowledge. A business skill pairs step-by-step instructions with the context an agent needs to execute a specific company procedure — renewal readiness, escalation protocols, onboarding workflows, compliance checks.</p>
<p>The key architectural decisions:</p>
<ul>
<li><strong>Define once, reuse everywhere.</strong> A skill defined by the sales operations team is accessible to any agent connected through the Work IQ MCP server, including agents built in Copilot Studio, third-party agents connected via A2A, and coding agents using the Work IQ plugin.</li>
<li><strong>Skills travel with the semantic model.</strong> When an agent invokes a business skill, it executes within the context of the organization&#8217;s semantic model — so a &#8220;renewal readiness&#8221; skill automatically understands what an &#8220;account,&#8221; &#8220;opportunity,&#8221; and &#8220;service case&#8221; mean in that organization.</li>
<li><strong>Skills replace scattered prompts.</strong> Instead of every agent builder copy-pasting the same &#8220;check for open service cases before recommending a renewal&#8221; instruction into their prompts, the skill encodes it once and keeps it current in one place.</li>
</ul>
<p>For RPA teams, this pattern will feel familiar — it&#8217;s the digital equivalent of a process definition document (PDD), but one that agents can actually execute rather than merely reference. Teams that have invested years in documenting SOPs for <a href="https://rpabotsworld.com/automation-anywhere-vs-uipath-agentic-procure-to-pay-2026/">procure-to-pay</a> or order-to-cash processes now have a path to expose those SOPs directly to agents.</p>
<h2>Governed Actions: How Agents Write Back Without Going Rogue</h2>
<p>The gap between an agent that can <em>read</em> business data and one that can <em>act</em> on it is where most enterprise agent projects stall. Work IQ addresses this with <strong>governed actions</strong> — a mechanism for agents to propose changes, route them through required approvals, and write updates back to source records.</p>
<p>The guardrails are layered:</p>
<ul>
<li><strong>User-scoped execution:</strong> Every action runs with the delegating user&#8217;s permissions, not the agent&#8217;s service credentials. An agent cannot escalate a service case if the user who invoked it doesn&#8217;t have permission to escalate cases.</li>
<li><strong>Approval routing:</strong> Changes that require sign-off are routed through the organization&#8217;s existing approval workflows. The agent proposes; the human approves; the system of record updates.</li>
<li><strong>System of record authority:</strong> The underlying system (Dynamics 365, a Power App, a third-party LOB app) remains authoritative. Agents don&#8217;t maintain their own copy of the data — they read from and write back to the same records everyone else uses.</li>
</ul>
<p>Examples of governed actions include escalating a case, reassigning opportunity ownership, updating a contract status, or creating a follow-up task — all without the user leaving their current workflow.</p>
<h2>The Rego Policy Engine: Enterprise Governance at the Tool Level</h2>
<p>This is the component that separates Work IQ from simpler &#8220;give the agent an API key and hope for the best&#8221; approaches. Every MCP tool invocation passes through a <strong>Rego-based policy engine</strong> — the same language used in the Cloud Native Computing Foundation&#8217;s <a href="https://www.openpolicyagent.org/" target="_blank" rel="noopener nofollow">Open Policy Agent (OPA)</a> project.</p>
<p>The policy engine evaluates each request against four dimensions:</p>
<ol>
<li><strong>Resource path:</strong> Which data is the agent trying to access?</li>
<li><strong>Request method:</strong> Is it a read or a write?</li>
<li><strong>User identity:</strong> Who delegated this action to the agent?</li>
<li><strong>Data content:</strong> What sensitivity labels or classifications apply to the target data?</li>
</ol>
<p>Administrators manage policies through the <strong>Microsoft 365 admin center → Agents → Tools</strong> interface. The default posture is conservative:</p>
<table>
<thead>
<tr>
<th>Policy Area</th>
<th>Default Setting</th>
</tr>
</thead>
<tbody>
<tr>
<td>Read operations (user&#8217;s own data, directory, SharePoint, Teams)</td>
<td>Enabled</td>
</tr>
<tr>
<td>Write actions (send email, create events/tasks)</td>
<td>Disabled</td>
</tr>
<tr>
<td>Create operations</td>
<td>Disabled</td>
</tr>
<tr>
<td>Partial update / Replace / Delete</td>
<td>Disabled</td>
</tr>
<tr>
<td>Deep paging / Continuation tokens</td>
<td>Disabled</td>
</tr>
</tbody>
</table>
<p>This means agents are read-only by default. An administrator must explicitly enable write actions before agents can send emails, create calendar events, or update records — a design choice that&#8217;s cautious to the point of friction, but defensible given the <a href="https://rpabotsworld.com/microsoft-copilot-autopilot-code-home-agentic-enterprise-guide/">always-on nature of Copilot Autopilot agents</a> that never stop running.</p>
<p>One critical caveat that Microsoft&#8217;s partners have flagged: <strong>Rego policies can&#8217;t fix pre-existing overbroad permissions.</strong> If SharePoint permissions are already too loose, Work IQ inherits that looseness. The recommended pre-deployment checklist includes a permissions audit, Purview sensitivity label review, Entra Conditional Access review, and sharing-settings lockdown before enabling agents on production data.</p>
<h2>Agent 365: The Governance Umbrella</h2>
<p>Work IQ agents operate under <strong>Microsoft Agent 365</strong>, which reached GA on May 1, 2026. Agent 365 provides the organizational controls that IT teams need before they&#8217;ll approve agents in production:</p>
<ul>
<li><strong>Agent registry</strong> — a central inventory of every agent running in the organization</li>
<li><strong>Visual topology map</strong> — see which agents connect to which data sources and tools</li>
<li><strong>Unmanaged agent discovery</strong> — detect local agents running outside the governed platform</li>
<li><strong>Security integration</strong> — Defender, Entra, Intune, and Purview protections extend to agents</li>
<li><strong>Agent identity</strong> — each Copilot Autopilot agent gets its own Entra identity, so actions are traceable to the specific agent that performed them</li>
</ul>
<p>For teams building on the <a href="https://rpabotsworld.com/microsoft-agent-framework-1-0-foundry-agentic-ai-architect-guide/">Microsoft Agent Framework 1.0 and Foundry Agent Service</a>, Agent 365 is the governance wrapper that makes those agents enterprise-acceptable.</p>
<h2>Copilot Studio Integration: Connecting Custom Agents to Work IQ</h2>
<p>For low-code builders and RPA teams already invested in the Power Platform, the Copilot Studio integration is the most practical entry point. The integration works through Work IQ MCP servers that can be enabled directly within a Copilot Studio agent configuration.</p>
<p>The available MCP servers within Copilot Studio include:</p>
<table>
<thead>
<tr>
<th>MCP Server</th>
<th>Capabilities</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Mail</strong></td>
<td>Draft or reply to emails, summarize threads, delete messages</td>
</tr>
<tr>
<td><strong>Calendar</strong></td>
<td>Create/update/retrieve Outlook events, manage invites, check availability</td>
</tr>
<tr>
<td><strong>Copilot</strong></td>
<td>Find and search Microsoft 365 content</td>
</tr>
<tr>
<td><strong>Teams</strong></td>
<td>Manage chat messages, create channels, add members</td>
</tr>
<tr>
<td><strong>Word</strong></td>
<td>Create/read documents, add or reply to comments</td>
</tr>
<tr>
<td><strong>OneDrive</strong></td>
<td>Manage personal files and folders (up to 5 MB per file)</td>
</tr>
<tr>
<td><strong>SharePoint</strong></td>
<td>Manage files, folders, and site administration (up to 5 MB per file)</td>
</tr>
<tr>
<td><strong>User</strong></td>
<td>Retrieve profile information for people in the organization</td>
</tr>
</tbody>
</table>
<p>A practical demo from ForwardForever showed a <strong>Major Incident Assist Agent</strong> built entirely with Work IQ MCP servers and no custom topics or knowledge sources. The agent could summarize an email thread to extract an incident number and SLA, find a 30-minute slot for an on-call team and create a Teams meeting with the incident link, and search Microsoft 365 for documents on the relevant issue type — all in a single conversational turn. Adding a ServiceNow MCP server to the same agent let it pull incident status, priority, and assignee from the ITSM system alongside the M365 context.</p>
<h2>Licensing and Cost Model</h2>
<p>Work IQ uses <strong>usage-based billing through Copilot Credits</strong>, which represents a significant shift from seat-based licensing:</p>
<ul>
<li><strong>Copilot-licensed users</strong> get Work IQ in all Copilot experiences and agents at no additional per-call cost</li>
<li><strong>Custom and third-party agents</strong> are billed consumptively — you pay for what agents use</li>
<li><strong>Users without a Copilot license</strong> can still use Work IQ through agents, billed consumptively</li>
<li><strong>Tools APIs</strong> have fixed costs; <strong>Chat and Context APIs</strong> are variable</li>
</ul>
<p>The consumption model means adoption isn&#8217;t gated by buying seats, but it also means high-frequency agents need careful unit-economics modeling. A customer service agent that processes thousands of tickets per day will generate a meaningful bill through cumulative tool invocations — the token savings from Work IQ&#8217;s pre-grounding (Microsoft claims approximately 80% reduction in token usage versus conventional retrieval) only offset costs when prompts would otherwise be very large.</p>
<p>Cost management controls are available in the Microsoft 365 admin center, and a usage-based billing plan in Copilot Studio with an Azure subscription and resource group is required for setup.</p>
<h2>How Work IQ Compares to Alternative Agent Platforms</h2>
<p>The enterprise agent platform market is consolidating around a few major stacks. Here&#8217;s how Work IQ&#8217;s approach compares:</p>
<table>
<thead>
<tr>
<th>Dimension</th>
<th>Microsoft Work IQ</th>
<th>Salesforce Agentforce</th>
<th>Google Vertex AI Agent Builder</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Data grounding</strong></td>
<td>M365 workplace graph + Dynamics 365 + Power Platform via shared semantic model</td>
<td>Salesforce Data Cloud unified profile</td>
<td>Vertex AI Search + BigQuery + enterprise data connectors</td>
</tr>
<tr>
<td><strong>Protocol support</strong></td>
<td>MCP + A2A (agent-to-agent) + REST</td>
<td>Proprietary Topics/Actions + emerging MCP</td>
<td>Vertex Extensions + A2A + MCP</td>
</tr>
<tr>
<td><strong>Governance</strong></td>
<td>Rego/OPA policy engine, user-scoped execution, Agent 365 registry</td>
<td>Einstein Trust Layer, data masking, audit trails</td>
<td>IAM-based, VPC-SC, Model Armor</td>
</tr>
<tr>
<td><strong>Low-code builder</strong></td>
<td>Copilot Studio</td>
<td>Agent Builder (Agentforce Studio)</td>
<td>Agent Builder (Vertex console)</td>
</tr>
<tr>
<td><strong>Pricing model</strong></td>
<td>Consumption-based (Copilot Credits)</td>
<td>Per-conversation pricing</td>
<td>Pay-per-use (Vertex pricing)</td>
</tr>
<tr>
<td><strong>Competitive moat</strong></td>
<td>Owns both productivity (M365) and business apps (Dynamics 365)</td>
<td>Owns the CRM data layer</td>
<td>Owns the data warehouse and ML infrastructure</td>
</tr>
</tbody>
</table>
<p>Microsoft&#8217;s structural advantage is that it owns both the productivity layer (Outlook, Teams, SharePoint) and the business applications layer (Dynamics 365, Power Platform). An agent that can reason across a customer&#8217;s email history, their CRM record, their service tickets, and their ERP data — all under one permission model — has a context advantage that competitors need partnerships and integrations to match.</p>
<p>The <a href="https://rpabotsworld.com/forrester-wave-adaptive-process-orchestration-q3-2026-guide/">Forrester Wave for Adaptive Process Orchestration (Q3 2026)</a> reflects this shift: the analysts now evaluate vendors on how well they connect agent reasoning to live business data, not just how well they execute predefined workflows.</p>
<h2>What This Means for RPA and Automation Teams</h2>
<p>If you&#8217;re running an RPA Center of Excellence on the Microsoft stack, Work IQ changes the calculation for what should be a bot versus what should be an agent:</p>
<h3>When to keep your existing RPA bots</h3>
<p>Deterministic processes that follow fixed rules — invoice processing with known templates, data entry between systems with rigid field mappings, scheduled report generation — are still better served by traditional RPA. Work IQ adds nothing to a bot that doesn&#8217;t need to <em>understand</em> business context.</p>
<h3>When to build Work IQ agents</h3>
<p>Processes that require judgment, context across multiple systems, or interaction with unstructured data are where Work IQ agents outperform bots. Sales renewal assessments that need CRM + service + email context. Employee onboarding coordination across HR, IT, and facilities. Procurement decisions that need to weigh contract terms against supplier communication history. These are the use cases where a bot would need hundreds of rules to approximate what a Work IQ-grounded agent handles natively.</p>
<h3>The hybrid architecture</h3>
<p>The pragmatic path for most teams is hybrid: Work IQ agents handle the judgment-heavy orchestration layer, and existing <a href="https://rpabotsworld.com/automation-anywhere-vs-uipath-agentic-procure-to-pay-2026/">RPA bots handle the deterministic execution</a> underneath. An agent decides <em>what</em> needs to happen based on context; a bot executes the specific system transactions. The Power Platform&#8217;s existing connectors and <a href="https://rpabotsworld.com/microsoft-copilot-autopilot-code-home-agentic-enterprise-guide/">Power Automate desktop flows</a> bridge the two worlds.</p>
<h2>Production Readiness Checklist</h2>
<p>Before deploying Work IQ agents to production, walk through this pre-flight sequence:</p>
<table>
<thead>
<tr>
<th>#</th>
<th>Check</th>
<th>Why It Matters</th>
</tr>
</thead>
<tbody>
<tr>
<td>1</td>
<td>Audit SharePoint and OneDrive permissions</td>
<td>Work IQ inherits existing permissions — overbroad access becomes overbroad agent access</td>
</tr>
<tr>
<td>2</td>
<td>Deploy Purview sensitivity labels</td>
<td>Labels travel with data and are enforced in agent working storage</td>
</tr>
<tr>
<td>3</td>
<td>Review Entra Conditional Access policies</td>
<td>Agents authenticate as users — CA policies must account for agent-delegated sessions</td>
</tr>
<tr>
<td>4</td>
<td>Provision the Work IQ service principal</td>
<td>One-time setup: <code>az ad sp create --id fdcc1f02-fc51-4226-8753-f668596af7f7</code></td>
</tr>
<tr>
<td>5</td>
<td>Configure Rego policies in M365 admin center</td>
<td>Default is read-only — explicitly enable only the write actions your agents need</td>
</tr>
<tr>
<td>6</td>
<td>Define business skills for core processes</td>
<td>Skills encode institutional knowledge once — don&#8217;t replicate it across agent prompts</td>
</tr>
<tr>
<td>7</td>
<td>Set up Copilot Credits billing and cost alerts</td>
<td>Consumption-based billing needs unit-economics modeling for high-frequency agents</td>
</tr>
<tr>
<td>8</td>
<td>Register agents in Agent 365</td>
<td>Unregistered agents are invisible to IT governance</td>
</tr>
</tbody>
</table>
<h2>Timeline: What&#8217;s Live Now and What&#8217;s Coming</h2>
<table>
<thead>
<tr>
<th>Date</th>
<th>Milestone</th>
</tr>
</thead>
<tbody>
<tr>
<td>June 2, 2026</td>
<td>Microsoft IQ umbrella announced at Build 2026</td>
</tr>
<tr>
<td>June 16, 2026</td>
<td>Work IQ APIs reach GA — A2A, redesigned MCP server, REST API</td>
</tr>
<tr>
<td>September 25, 2026</td>
<td>Work IQ grounding for Dynamics 365 and Power Platform announced</td>
</tr>
<tr>
<td>September 30, 2026</td>
<td>Dynamics 365 Sales + Customer Service + Power Apps grounding enters public preview</td>
</tr>
<tr>
<td>October 2026</td>
<td>Public preview rollout continues; third-party LOB app support</td>
</tr>
<tr>
<td>Late October 2026</td>
<td>Dynamics 365 Finance and Operations joins the preview — ERP data enters agent context</td>
</tr>
</tbody>
</table>
<h2>Frequently Asked Questions</h2>
<h3>Do I need a Microsoft 365 Copilot license to use Work IQ?</h3>
<p>Not necessarily. Work IQ API access is consumption-based and independent of Copilot licensing. Users without a Copilot license can access Work IQ through agents, billed consumptively via Copilot Credits. However, Copilot-licensed users get Work IQ included in all Copilot experiences at no additional per-call cost, and using Work IQ MCP servers within Copilot Studio requires a Microsoft 365 Copilot license.</p>
<h3>Can I connect non-Microsoft agents (LangChain, CrewAI, AutoGen) to Work IQ?</h3>
<p>Yes. The Work IQ MCP server follows the Model Context Protocol standard, and the A2A protocol enables agent-to-agent delegation. Any MCP-compatible agent framework can connect to the Work IQ endpoint at <code>https://workiq.svc.cloud.microsoft/mcp</code> using Entra ID delegated authentication. Microsoft&#8217;s Foundry Agent Service also supports hosted agents that can call Claude, GPT, Gemini, or other models.</p>
<h3>How does Work IQ handle multi-tenant organizations?</h3>
<p>For parent/child tenant configurations, register the app as <code>AzureADMultipleOrgs</code>. The token&#8217;s <code>iss</code> (issuer) must match the signed-in user&#8217;s tenant, or requests fail with a <code>400 AuthenticationError</code>. Each tenant maintains its own semantic model, business skills, and Rego policies.</p>
<h3>What happens when an agent action is blocked by policy?</h3>
<p>Denied requests return a policy error, typically <code>403 Forbidden</code>. Agents should handle these responses gracefully and inform users that governance — not a bug — blocked the action. An operation that works in a dev tenant may fail in a production tenant with different policies configured.</p>
<h3>Can Work IQ replace my existing custom RAG pipelines for M365 data?</h3>
<p>For Microsoft 365 data, Work IQ is designed to supersede custom RAG pipelines. Microsoft&#8217;s stated advantages include permission-aware retrieval without permission drift, real-time data freshness without manual reindexing, and approximately 80% reduction in token usage compared to conventional retrieval approaches. However, for data outside the Microsoft ecosystem, you&#8217;ll still need custom retrieval — Work IQ doesn&#8217;t index Salesforce records or SAP transactions.</p>
<h2>Key Takeaways</h2>
<ul>
<li><strong>Work IQ is Microsoft&#8217;s answer to the agent context gap</strong> — the reason most enterprise agent pilots stall is not model capability but business data grounding, and Work IQ provides a shared semantic model across M365, Dynamics 365, and Power Platform.</li>
<li><strong>The MCP server design (12 tools, path-based routing) is architecturally significant</strong> — it reduces tool-selection complexity for LLMs and scales without adding tools as new workloads come online.</li>
<li><strong>Rego-based governance isn&#8217;t optional</strong> — the OPA policy engine evaluates every tool invocation, and the default posture is read-only. This is conservative by design and reflects lessons from early agent deployments where agents exceeded intended boundaries.</li>
<li><strong>Business skills are the enterprise equivalent of process definition documents</strong> — define once, reuse across all connected agents, and keep institutional knowledge in one place rather than scattered across prompts.</li>
<li><strong>The October 2026 preview extending to Dynamics 365 and Power Platform is the real inflection point</strong> — Work IQ grounding M365 productivity data was useful; grounding CRM and ERP data makes it transformational for business process automation.</li>
<li><strong>RPA teams should plan hybrid architectures</strong> — Work IQ agents for judgment-heavy orchestration, existing bots for deterministic execution, connected through Power Automate flows.</li>
</ul>
<h2>References</h2>
<ol>
<li>Microsoft Dev Blog, &#8220;Work IQ: Production-Ready Intelligence for Every Agent,&#8221; June 2, 2026 — <a href="https://devblogs.microsoft.com/microsoft365dev/work-iq-production-ready-intelligence-for-every-agent/" target="_blank" rel="noopener nofollow">devblogs.microsoft.com</a></li>
<li>Microsoft Learn, &#8220;Work IQ MCP Server Overview,&#8221; 2026 — <a href="https://learn.microsoft.com/en-us/microsoft-365-copilot/extensibility/work-iq/mcp/overview" target="_blank" rel="noopener nofollow">learn.microsoft.com</a></li>
<li>ERP Today, &#8220;Microsoft Opens Work IQ Preview for Dynamics 365 and Power Platform Data,&#8221; October 2026 — <a href="https://erp.today/microsoft-opens-work-iq-preview-for-dynamics-365-and-power-platform-data/" target="_blank" rel="noopener nofollow">erp.today</a></li>
<li>Cloud Wars, &#8220;Microsoft&#8217;s Work IQ Gets Down to Business,&#8221; September 28, 2026 — <a href="https://cloudwars.com/ai/microsofts-work-iq-gets-down-to-business-and-thats-what-makes-agents-useful/" target="_blank" rel="noopener nofollow">cloudwars.com</a></li>
<li>EPC Group, &#8220;Microsoft Build 2026: Work IQ, Microsoft IQ, Agent 365 Context Layer,&#8221; 2026 — <a href="https://www.epcgroup.net/blog/microsoft-build-2026-work-iq-microsoft-iq-agent-365-context-layer-2026" target="_blank" rel="noopener nofollow">epcgroup.net</a></li>
<li>ForwardForever, &#8220;Extending Copilot Studio Agents with Work IQ Mail, Calendar, and Copilot Tools in Action,&#8221; 2026 — <a href="https://forwardforever.com/extending-copilot-studio-agents-with-work-iq-mail-calendar-and-copilot-tools-in-action/" target="_blank" rel="noopener nofollow">forwardforever.com</a></li>
<li>Candede, &#8220;Integrate Work IQ MCP,&#8221; 2026 — <a href="https://candede.com/articles/integrate-work-iq-mcp" target="_blank" rel="noopener nofollow">candede.com</a></li>
<li>AI Automation Global, &#8220;Microsoft IQ Platform: Work IQ APIs, Enterprise Agent Context 2026&#8221; — <a href="https://aiautomationglobal.com/blog/microsoft-iq-platform-work-iq-apis-enterprise-agent-context-2026" target="_blank" rel="noopener nofollow">aiautomationglobal.com</a></li>
<li>Microsoft, &#8220;Announcing the New Work IQ APIs,&#8221; June 2, 2026 — <a href="https://microsoft.com/en-us/microsoft-365/blog/2026/06/02/announcing-the-new-work-iq-apis" target="_blank" rel="noopener nofollow">microsoft.com</a></li>
<li>Open Policy Agent documentation — <a href="https://www.openpolicyagent.org/" target="_blank" rel="noopener nofollow">openpolicyagent.org</a></li>
</ol>
<p><!-- SEO METADATA
Title: Microsoft Work IQ and MCP Server: Complete Guide for Agentic AI Architects (2026)
Meta Description: Microsoft Work IQ grounds Copilot agents in Dynamics 365 and Power Platform data via MCP. Architecture, Rego policies, business skills, and production checklist.
Slug: microsoft-work-iq-mcp-server-dynamics-365-agentic-ai-guide
Target Keyword: Microsoft Work IQ
Secondary Keywords: Work IQ MCP server, Copilot agents Dynamics 365, Microsoft IQ agentic AI
Search Intent: Informational / Technical deep-dive
--></p>
<p><!-- SOCIAL MEDIA POST (X/Twitter)
Microsoft Work IQ just hit public preview with Dynamics 365 + Power Platform grounding.

12 generic MCP tools replace hundreds of Graph API calls. Rego-based policy engine governs every invocation. Business skills codify your SOPs for agents.

The complete guide for agentic AI architects &#x1f447;
https://rpabotsworld.com/microsoft-work-iq-mcp-server-dynamics-365-agentic-ai-guide/
--></p>
<p><!-- LINKEDIN POST
Microsoft just shipped something that matters more than another model upgrade: Work IQ — the intelligence layer that grounds Copilot and AI agents in your actual business data.

Here's what changed in October 2026:

→ Dynamics 365 Sales, Customer Service, and Power Apps data now feed into the same context layer as email, Teams, and files
→ 12 generic MCP tools replace hundreds of granular API operations — agents pick paths, not tools
→ A Rego-based policy engine (Open Policy Agent) evaluates every single tool invocation against resource path, method, identity, and data sensitivity
→ Business skills let you define a company procedure once and share it across every connected agent
→ Late October adds Finance and Operations — ERP data enters agent context

Why this matters for RPA teams:

The hardest part of enterprise agent adoption was never the model — it was grounding agents in the data they need to make good decisions. Work IQ bridges the gap between "agents that can chat" and "agents that understand your business."

For automation architects, the play is hybrid: Work IQ agents handle judgment-heavy orchestration, existing RPA bots handle deterministic execution underneath.

Full technical deep-dive with architecture, MCP server reference, governance model, and production readiness checklist:
https://rpabotsworld.com/microsoft-work-iq-mcp-server-dynamics-365-agentic-ai-guide/

#AgenticAI #Microsoft #WorkIQ #CopilotAgents #RPA #Automation #Dynamics365 #PowerPlatform #MCP
--></p>
<p><!-- IMAGE SUGGESTIONS
1. Architecture diagram showing the four layers of Work IQ (Chat → Context → Tools → Workspaces) with Dynamics 365, Power Platform, and M365 data sources feeding in, and Copilot Studio / third-party agents consuming context through MCP
2. Comparison table visual: Microsoft Work IQ vs Salesforce Agentforce vs Google Vertex AI Agent Builder across data grounding, protocols, governance, and pricing
3. Flowchart showing a governed action lifecycle: Agent proposes → Rego policy evaluates → Approval routing → System of record updates
--></p>
]]></content:encoded>
					
					<wfw:commentRss>https://rpabotsworld.com/microsoft-work-iq-mcp-server-dynamics-365-agentic-ai-guide/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>Oracle AI Agent Studio for Industry Applications: The Complete Guide for Agentic AI Architects (2026)</title>
		<link>https://rpabotsworld.com/oracle-ai-agent-studio-industry-applications-guide/</link>
					<comments>https://rpabotsworld.com/oracle-ai-agent-studio-industry-applications-guide/#respond</comments>
		
		<dc:creator><![CDATA[Satish Prasad]]></dc:creator>
		<pubDate>Thu, 22 Oct 2026 02:38:34 +0000</pubDate>
				<category><![CDATA[RPA & Bot Automation]]></category>
		<category><![CDATA[Agentic AI & AI Automation]]></category>
		<category><![CDATA[AI Agents & Frameworks]]></category>
		<category><![CDATA[MCP]]></category>
		<guid isPermaLink="false">https://rpabotsworld.com/?p=32461</guid>

					<description><![CDATA[Oracle AI Agent Studio for Industry Applications brings no-code agent building, MCP connectivity, and multi-agent orchestration to 8 industry verticals.]]></description>
										<content:encoded><![CDATA[
<p class="wp-block-paragraph">On October 21, 2026, Oracle quietly did something none of the usual agentic AI suspects — not Microsoft, not Salesforce, not Google — have managed yet: it shipped a no-code agent-building platform that speaks the language of specific industries out of the box. Not &#8220;enterprise&#8221; in the abstract. Construction. Utilities. Hospitality. Government. Eight verticals, each with prebuilt agents, workflows, and templates wired into Oracle&#8217;s Industry Applications data layer, coordinated through Model Context Protocol (MCP) servers, and backed by a bring-your-own-model architecture that supports OpenAI, Anthropic, Google, Meta, and xAI models through OCI Enterprise AI.</p>



<p class="wp-block-paragraph">For RPA and automation architects who have spent the last 18 months watching every enterprise vendor launch an &#8220;agent platform&#8221; that still requires six months of custom integration before it touches real business data, Oracle AI Agent Studio for Industry Applications represents a different thesis: that the real bottleneck in enterprise agentic AI is not the agent runtime — it is the distance between the agent and the industry-specific processes, compliance rules, and data models it needs to be useful.</p>



<p class="wp-block-paragraph">This guide breaks down what Oracle announced, how the Industry Applications extension builds on 19 months of iteration on the Fusion AI Agent Studio, where it fits in the competitive landscape alongside <a href="https://rpabotsworld.com/salesforce-agentforce-job-ready-agents-long-horizon-runtime-guide-3/">Salesforce Agentforce</a>, Microsoft Copilot Studio, and <a href="https://rpabotsworld.com/google-gemini-enterprise-agent-platform-guide/">Google&#8217;s Gemini Enterprise Agent Platform</a>, and what it means for your automation practice.</p>



<h2 class="wp-block-heading">The Oracle Agentic AI Timeline: From 50 Agents to 1,000+ in 19 Months</h2>



<p class="wp-block-paragraph">Oracle&#8217;s agent story did not start with this announcement. To understand where the Industry Applications studio sits, you need the full timeline:</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Date</th><th>Milestone</th><th>Significance</th></tr></thead><tbody><tr><td>September 2024</td><td>Oracle announces 50+ AI agents for Fusion Cloud Applications</td><td>First wave: finance, HR, supply chain, CX. Agents pair Cohere-hosted LLMs with RAG over enterprise data. No additional cost to Fusion customers.</td></tr><tr><td>March 2025</td><td>AI Agent Studio launches at CloudWorld London</td><td>Customers and partners can build, test, and deploy custom agents within Fusion apps for the first time. No-code authoring with natural-language instructions.</td></tr><tr><td>October 2025</td><td>AI Agent Marketplace opens</td><td>Partners offer Oracle-validated agent templates, connectors, and full agentic applications — creating a network effect.</td></tr><tr><td>March 2026</td><td>Agentic Applications Builder added</td><td>Expanded the studio with visual workflow tools, deeper orchestration, and MCP server connectivity for tool reuse.</td></tr><tr><td>April 2026</td><td>Steve Miranda confirms 1,000+ agents</td><td>Oracle EVP of Applications Development reveals the agent count grew from ~50 to over 1,000, with 22 &#8220;Fusion Agentic Applications&#8221; — packaged multi-agent workflows.</td></tr><tr><td>July 2026</td><td>AI-native builder (&#8220;Fusion Agentic Applications&#8221;) released</td><td>New software category: systems of specialized agents taking outcome-driven action. Developers can build agents in VS Code with Git workflows using OpenAI Codex and Claude Code.</td></tr><tr><td>October 21, 2026</td><td><strong>AI Agent Studio for Industry Applications</strong></td><td>Extends the platform beyond Fusion into industry-specific applications across 8 verticals. Cross-application orchestration, MCP-server tool connectivity, and multi-agent team coordination.</td></tr></tbody></table></figure>



<p class="wp-block-paragraph">That timeline matters because Oracle&#8217;s approach is incremental and accretive — each release extends the surface area where agents can operate, without replacing what came before. The Industry Applications studio is not a separate product. It is the same authoring, testing, and governance environment that 80,000+ certified professionals already know from the Fusion studio, extended to reach industry-specific data, processes, and compliance requirements.</p>



<figure class="wp-block-image size-full"><img fetchpriority="high" decoding="async" width="2200" height="1558" src="https://rpabotsworld.com/wp-content/uploads/2026/10/oracle-ai-agent-studio-industry-applications-guide-architecture-diagram.png" alt="Oracle AI Agent Studio for Industry Applications: The Complete Guide for Agentic AI Architects (2026) 1" class="wp-image-32463" title="Oracle AI Agent Studio for Industry Applications: The Complete Guide for Agentic AI Architects (2026) 1"></figure>



<h2 class="wp-block-heading">What Oracle AI Agent Studio for Industry Applications Actually Is</h2>



<p class="wp-block-paragraph">At its core, this is a no-code and low-code environment for creating, testing, orchestrating, deploying, and managing AI agents and workflows that operate on industry-specific business processes. It works across three Oracle layers simultaneously:</p>



<ul class="wp-block-list">
<li><strong>Oracle Industry Applications</strong> — vertical solutions for construction, retail, financial services, restaurants, government, utilities, communications, and hospitality</li>



<li><strong>Oracle Fusion Cloud Applications</strong> — the horizontal ERP, HCM, SCM, and CX suite</li>



<li><strong>Oracle Cloud Infrastructure (OCI)</strong> — the underlying compute, AI models, and data services</li>
</ul>



<p class="wp-block-paragraph">The key architectural insight is cross-application orchestration: agents can coordinate multi-step workflows that span both industry-specific and Fusion applications without requiring data movement or point-to-point integrations. A construction project manager&#8217;s agent can pull schedule data from the industry application, cross-reference it with procurement workflows in Fusion SCM, and flag supply-chain risks that affect milestone delivery — all within the same governed, auditable execution environment.</p>



<h2 class="wp-block-heading">Six Capabilities That Matter for Automation Architects</h2>



<h3 class="wp-block-heading">1. Prebuilt Industry Foundations</h3>



<p class="wp-block-paragraph">Unlike horizontal agent platforms where every deployment starts from a blank canvas, Oracle ships prebuilt industry agents, workflows, templates, and reusable tools for each supported vertical. These are not demos — they are production-ready starting points that encode industry-specific process knowledge:</p>



<ul class="wp-block-list">
<li><strong>Engineering and construction:</strong> Agents that identify risks to schedule milestones, find related field issues across projects, and propose mitigation steps — the kind of cross-referencing that typically requires a senior project controls specialist</li>



<li><strong>Retail:</strong> Agents for product image analysis, catalog management, and product information enrichment — connecting merchandising data with supply chain visibility</li>



<li><strong>Financial services:</strong> Agents handling payments processing, affordability analysis, and customer preference modeling — built within the regulatory guardrails financial institutions require</li>
</ul>



<p class="wp-block-paragraph">For RPA practitioners accustomed to the &#8220;80% of project time is spent understanding the process&#8221; reality, prebuilt industry foundations address the most expensive part of automation delivery: domain knowledge acquisition.</p>



<h3 class="wp-block-heading">2. Natural-Language Authoring with Visual Workflow Editing</h3>



<p class="wp-block-paragraph">Agent creation uses a hybrid approach. Business users define agent roles, goals, tools, policies, and operating context through natural-language instructions — describe what the agent should do, not how it should do it. Developers who need more control get visual workflow editing that shows the agent&#8217;s decision tree, tool calls, and escalation paths graphically.</p>



<p class="wp-block-paragraph">This dual-mode authoring maps directly to how enterprise automation teams actually work: business analysts define requirements in business language, and solution architects translate them into executable workflows. Oracle&#8217;s studio collapses that handoff into a single environment.</p>



<h3 class="wp-block-heading">3. MCP-Based Tool Connectivity</h3>



<p class="wp-block-paragraph">This is where Oracle makes a bet that will resonate with anyone following the <a href="https://rpabotsworld.com/appian-agentic-automation-mcp-data-fabric-gartner-guide/">Model Context Protocol movement in enterprise automation</a>. The studio uses MCP servers to connect agents to approved Oracle Industry Application tools and capabilities. Those MCP servers are reusable across multiple use cases — build the connection once, and every agent in the organization can discover and use the tool through the protocol.</p>



<p class="wp-block-paragraph">MCP is the open standard (originated by Anthropic) that defines how AI agents discover and interact with external tools and data sources. Oracle&#8217;s adoption of it — rather than building a proprietary tool-connectivity layer — signals a recognition that enterprise agent platforms will live or die on interoperability. The same protocol is used by Claude, by <a href="https://rpabotsworld.com/appian-agentic-automation-mcp-data-fabric-gartner-guide/">Appian&#8217;s agentic platform</a>, and by an expanding ecosystem of open-source and commercial MCP servers.</p>



<h3 class="wp-block-heading">4. Multi-Agent Teams and Cross-Application Orchestration</h3>



<p class="wp-block-paragraph">The studio supports &#8220;teams of agents&#8221; — multiple specialized agents coordinating on a multi-step business process that spans applications. This is not just running agents in parallel; it is orchestrated workflow coordination where Agent A&#8217;s output becomes Agent B&#8217;s input, with the orchestration layer managing sequencing, error handling, and human-in-the-loop checkpoints.</p>



<p class="wp-block-paragraph">Oracle&#8217;s differentiation here is that the orchestration happens across the full Oracle stack — Industry Applications, Fusion, and OCI services — with native data access at every layer. Competitors typically orchestrate agents within a single application boundary (Salesforce Agentforce within Salesforce, Copilot Studio within Microsoft 365) and treat cross-application workflows as an integration problem.</p>



<h3 class="wp-block-heading">5. Model Choice via OCI Enterprise AI</h3>



<p class="wp-block-paragraph">Oracle does not lock you into a single foundation model. Through OCI Enterprise AI, agents can select from supported generative AI models from OpenAI, Anthropic, Google, Meta, and xAI — and the choice can be made per agent or per use case. Need a cost-efficient model for routine document classification? Use a smaller model. Need state-of-the-art reasoning for complex financial analysis? Switch to a frontier model. The platform abstracts the model layer so the agent logic remains the same regardless of which LLM powers it.</p>



<p class="wp-block-paragraph">This multi-model flexibility is a strategic advantage in an environment where model capabilities, pricing, and availability are changing quarterly. It also addresses a common enterprise concern: vendor lock-in at the model layer.</p>



<h3 class="wp-block-heading">6. Enterprise-Grade Governance and Monitoring</h3>



<p class="wp-block-paragraph">Oracle built governance into the architecture rather than bolting it on after the fact — an approach that reflects the company&#8217;s decades of experience with enterprise audit and compliance requirements:</p>



<ul class="wp-block-list">
<li><strong>Guardrails</strong> keep agents within configured permissions, policies, and organizational roles</li>



<li><strong>Access controls</strong> ensure an agent cannot see data that its user cannot see — the same role-based security that governs the underlying Oracle applications</li>



<li><strong>Audit trails</strong> record every agent action for compliance and debugging</li>



<li><strong>Testing and debugging tools</strong> let you trace workflow execution, inspect inputs and outputs, evaluate results, and retest changes before production deployment</li>



<li><strong>Monitoring dashboards</strong> track execution metrics, performance, reliability, usage, and AI-unit consumption — critical for cost management at scale</li>
</ul>



<p class="wp-block-paragraph">For architects coming from the RPA world, this governance model will feel familiar — it mirrors the control-tower approach that platforms like UiPath Orchestrator and <a href="https://rpabotsworld.com/automation-anywhere-vs-uipath-agentic-procure-to-pay-2026/">Automation Anywhere&#8217;s Control Room</a> use for bot management, extended to the more complex reality of LLM-powered agents that make probabilistic decisions.</p>



<h2 class="wp-block-heading">How It Compares: Oracle vs. the Enterprise Agent Platform Landscape</h2>



<p class="wp-block-paragraph">Oracle AI Agent Studio for Industry Applications enters a crowded field. Here is how it stacks up against the four other major enterprise agent platforms that RPA and automation architects are evaluating in late 2026:</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Dimension</th><th>Oracle AI Agent Studio (Industry)</th><th>Microsoft Copilot Studio</th><th>Salesforce Agentforce</th><th>Google Gemini Enterprise Agent Platform</th><th>Appian Agentic Automation</th></tr></thead><tbody><tr><td><strong>Industry specialization</strong></td><td>8 industry verticals with prebuilt agents and templates</td><td>Horizontal (M365/Dynamics-centric)</td><td>CRM/Sales/Service-centric; industry clouds exist but are separate</td><td>Horizontal (GCP-centric)</td><td>Horizontal with Data Fabric</td></tr><tr><td><strong>Agent authoring</strong></td><td>No-code (NL instructions) + low-code (visual workflows) + pro-code (VS Code/Git)</td><td>No-code + low-code (Power Automate)</td><td>Low-code (Agent Builder) + Apex pro-code</td><td>Code-first (ADK/Python) + Vertex AI console</td><td>Low-code (Process Modeler) + SAIL</td></tr><tr><td><strong>Multi-agent orchestration</strong></td><td>Native multi-agent teams across Industry + Fusion apps</td><td>Agent-to-agent via shared memory and triggers</td><td>Multi-agent via Topics routing</td><td>Agent-to-agent via A2A protocol</td><td>Process-level orchestration</td></tr><tr><td><strong>Tool connectivity</strong></td><td>MCP servers (open standard)</td><td>Connectors + custom plugins</td><td>MuleSoft + Flow actions</td><td>Extensions + Function Calling + MCP (partial)</td><td>MCP + Data Fabric connectors</td></tr><tr><td><strong>Model flexibility</strong></td><td>Multi-vendor: OpenAI, Anthropic, Google, Meta, xAI</td><td>OpenAI (primary), some Azure AI models</td><td>Salesforce Einstein models + partner models</td><td>Gemini models (primary)</td><td>Multi-vendor via Private AI</td></tr><tr><td><strong>Marketplace</strong></td><td>AI Agent Marketplace (since Oct 2025)</td><td>Power Platform marketplace</td><td>AppExchange</td><td>Google Cloud Marketplace</td><td>Appian Community</td></tr><tr><td><strong>Data proximity</strong></td><td>Native access to Industry + Fusion + OCI data</td><td>Microsoft Graph + Dataverse</td><td>Salesforce Data Cloud</td><td>BigQuery + AlloyDB</td><td>Data Fabric (cross-system)</td></tr><tr><td><strong>Pricing signal</strong></td><td>No additional cost for eligible Fusion customers (Fusion studio); industry version TBD</td><td>Per-message consumption</td><td>Per-conversation pricing</td><td>GCP consumption-based</td><td>Platform license + AI units</td></tr></tbody></table></figure>



<h3 class="wp-block-heading">Oracle&#8217;s Competitive Edge: Data Proximity + Industry Context</h3>



<p class="wp-block-paragraph">The thesis Oracle is betting on — and it is a defensible one — is that an agent&#8217;s usefulness is directly proportional to how close it sits to the data and processes it needs to act on. As Eric Newmark, GVP and GM of Enterprise Business Applications at IDC, put it in the announcement: &#8220;Agents are only as useful as the data, tools, and approvals they can reach, and connecting all three is where most efforts stall.&#8221;</p>



<p class="wp-block-paragraph">For organizations already running Oracle Industry Applications, the studio eliminates the most expensive integration layer. The agent does not need a connector to reach the construction project schedule or the retail catalog — it is already inside the application that owns that data, with the same security model, the same audit trail, and the same compliance posture.</p>



<p class="wp-block-paragraph">The weakness is the mirror image of the strength: if you are not an Oracle shop, the Industry Applications studio offers you nothing. It is a platform-native play, not a platform-agnostic one. Organizations running SAP, Salesforce, or Microsoft Dynamics for their industry applications would need to evaluate those vendors&#8217; own agent platforms or look at cross-platform solutions like Appian.</p>



<h2 class="wp-block-heading">Architecture Deep Dive: How Agents Flow Through the Oracle Stack</h2>



<p class="wp-block-paragraph">Understanding the runtime architecture is essential for anyone planning to deploy agents in production. Here is how the layers connect:</p>



<p class="wp-block-paragraph"><strong>Layer 1 — OCI Enterprise AI (Model Layer):</strong> Foundation models from multiple providers (OpenAI, Anthropic, Google, Meta, xAI) run on OCI infrastructure. Private endpoints keep AI traffic on private networks — critical for regulated industries in financial services, government, and healthcare. Oracle Database 26ai provides AI Vector Search directly from SQL, supporting retrieval-augmented generation (RAG) without moving data to separate vector stores.</p>



<p class="wp-block-paragraph"><strong>Layer 2 — AI Agent Studio (Authoring and Orchestration Layer):</strong> This is where agents are built, tested, and managed. The studio provides the authoring canvas (natural language + visual workflows), the debugging tools (trace execution, inspect I/O), and the orchestration engine that coordinates multi-agent teams across applications. MCP servers plug into this layer, exposing Oracle Industry Application tools as discoverable, reusable capabilities.</p>



<p class="wp-block-paragraph"><strong>Layer 3 — Oracle Applications (Execution Layer):</strong> Agents execute actions within Oracle Industry Applications and Fusion Cloud Applications — reading data, triggering workflows, updating records, and escalating to human approvers. The agent inherits the calling user&#8217;s permissions, ensuring it cannot access data or perform actions beyond what the user is authorized to do.</p>



<p class="wp-block-paragraph"><strong>Layer 4 — Governance and Monitoring (Control Layer):</strong> Cross-cutting concerns — guardrails, audit logs, execution metrics, AI-unit consumption tracking — operate across all three layers above. This is Oracle&#8217;s analog to the &#8220;control tower&#8221; concept familiar from RPA platforms, adapted for the higher complexity of LLM-powered agents.</p>



<h2 class="wp-block-heading">Industry Use Cases: What You Can Actually Build</h2>



<h3 class="wp-block-heading">Engineering and Construction</h3>



<p class="wp-block-paragraph">Construction projects generate enormous volumes of data across scheduling systems, field reports, change orders, RFIs, and safety logs. Oracle&#8217;s prebuilt construction agents can cross-reference schedule milestones against active field issues — something that today requires a project controls team manually correlating data across multiple systems. The agent identifies which field issues threaten which milestones, proposes mitigation steps, and routes the analysis to the project manager for approval.</p>



<p class="wp-block-paragraph">For RPA architects who have built attended bots for construction firms, this represents the next evolution: instead of automating the data extraction step and leaving the analysis to humans, the agent handles both extraction and analysis, with the human reviewing the agent&#8217;s recommendations rather than raw data.</p>



<h3 class="wp-block-heading">Retail</h3>



<p class="wp-block-paragraph">Retail agents can manage product catalog workflows — analyzing product images, enriching product descriptions, and maintaining catalog consistency across channels. In a sector where a single retailer might manage hundreds of thousands of SKUs, even small efficiency gains in catalog management translate to meaningful revenue impact. The agent operates within Oracle Retail&#8217;s data model, so it has native access to pricing, inventory, and merchandising context.</p>



<h3 class="wp-block-heading">Financial Services</h3>



<p class="wp-block-paragraph">Financial services agents handle payments processing, affordability analysis, and customer preference modeling — all areas where accuracy, compliance, and auditability are non-negotiable. Oracle&#8217;s guardrail architecture is particularly relevant here: agents operate within the same role-based access controls and audit trail infrastructure that the underlying financial applications use, which simplifies the compliance conversation with regulators.</p>



<h3 class="wp-block-heading">Government, Utilities, and Beyond</h3>



<p class="wp-block-paragraph">Oracle also names government, utilities, communications, hospitality, and restaurants as supported verticals. While specific use cases for these sectors are not detailed in the initial announcement, the platform architecture — prebuilt agents and templates that encode industry-specific process knowledge, connected to industry application data through MCP servers — applies consistently across all eight verticals.</p>



<h2 class="wp-block-heading">What This Means for RPA and Automation Practitioners</h2>



<p class="wp-block-paragraph">If you are an RPA developer or solution architect evaluating Oracle AI Agent Studio for Industry Applications, here are the practical implications:</p>



<p class="wp-block-paragraph"><strong>If you are already in the Oracle ecosystem:</strong> This is the most direct path to agentic AI for your industry processes. The studio extends tools you may already know from the Fusion AI Agent Studio, with no additional learning curve for the authoring environment. Start with the prebuilt industry agents — they encode domain knowledge that would take weeks to capture manually — and customize from there.</p>



<p class="wp-block-paragraph"><strong>If you run a hybrid landscape:</strong> Oracle&#8217;s strength (data proximity within its own stack) becomes a limitation when your processes span Oracle and non-Oracle systems. MCP server support helps bridge some gaps — if the external system has an MCP server, agents can connect to it — but cross-platform orchestration is not Oracle&#8217;s primary design center. Evaluate whether <a href="https://rpabotsworld.com/forrester-wave-adaptive-process-orchestration-q3-2026-guide/">adaptive process orchestration platforms</a> better fit your cross-system reality.</p>



<p class="wp-block-paragraph"><strong>If you are comparing enterprise agent platforms:</strong> Oracle&#8217;s differentiator is industry depth. Microsoft, Salesforce, and Google all offer strong horizontal agent platforms, but none of them ship prebuilt agents for construction schedule risk analysis or retail catalog management out of the box. If your automation use cases are industry-specific and your data already lives in Oracle, the time-to-value argument is compelling.</p>



<p class="wp-block-paragraph"><strong>If you are an RPA consultant:</strong> Oracle&#8217;s 80,000+ certified AI professional base represents a significant addressable market, and the AI Agent Marketplace creates partnership opportunities. The combination of no-code authoring (for business-user enablement) and pro-code extensibility (for technical consultants) maps to the two-tier delivery model most consulting firms use.</p>



<h2 class="wp-block-heading">Availability, Pricing, and What to Watch</h2>



<p class="wp-block-paragraph">Oracle AI Agent Studio for Industry Applications is planned to be available within the next 12 months, initially in the United States, with expansion to all supported OCI Enterprise AI regions over time. The Fusion version of the studio has been available at no additional cost to eligible customers and partners since March 2025 — Oracle has not confirmed whether the same pricing applies to the Industry Applications extension.</p>



<p class="wp-block-paragraph">The monitoring dashboard&#8217;s tracking of &#8220;AI-unit consumption&#8221; suggests a usage-based component may be coming, but Oracle has not disclosed pricing details. Keep an eye on Oracle&#8217;s next CloudWorld event for clarity on the commercial model.</p>



<h3 class="wp-block-heading">What IDC, KPMG, and PwC Are Saying</h3>



<p class="wp-block-paragraph">The analyst and partner commentary around the launch is measured but positive:</p>



<ul class="wp-block-list">
<li><strong>Eric Newmark (IDC):</strong> &#8220;Agents are only as useful as the data, tools, and approvals they can reach, and connecting all three is where most efforts stall. [This platform] helps teams build, test, and manage agents within existing industry workflows. That is what turns adoption from a pilot into something practical, and something that scales.&#8221;</li>



<li><strong>Nick Yoko (KPMG):</strong> Called the platform a reflection of &#8220;an important shift toward more industry-specific AI&#8221; and emphasized that industry context is critical as organizations move from AI experimentation to value.</li>



<li><strong>Kevin Sullivan (PwC):</strong> Highlighted the combination of &#8220;industry context with testing, monitoring, and access controls to help build governance into workflows.&#8221;</li>
</ul>



<p class="wp-block-paragraph">The common thread across all three perspectives: the value is not in the agent runtime technology itself — it is in how tightly the agents are woven into the industry-specific processes, data, and governance structures that enterprises already have.</p>



<h2 class="wp-block-heading">Frequently Asked Questions</h2>



<h3 class="wp-block-heading">What is Oracle AI Agent Studio for Industry Applications?</h3>



<p class="wp-block-paragraph">It is a no-code and low-code platform for building, testing, orchestrating, and managing AI agents that work within Oracle&#8217;s industry-specific applications across eight verticals: engineering and construction, retail, financial services, restaurants, government, utilities, communications, and hospitality. It extends the existing Oracle AI Agent Studio for Fusion Applications.</p>



<h3 class="wp-block-heading">Does Oracle AI Agent Studio support multi-agent workflows?</h3>



<p class="wp-block-paragraph">Yes. The platform supports &#8220;teams of agents&#8221; — multiple specialized agents coordinating on multi-step business processes that span both Oracle Industry Applications and Oracle Fusion Cloud Applications. The orchestration engine manages sequencing, tool calls, and human-in-the-loop checkpoints.</p>



<h3 class="wp-block-heading">Which AI models does Oracle AI Agent Studio support?</h3>



<p class="wp-block-paragraph">Through OCI Enterprise AI, the studio supports generative AI models from OpenAI, Anthropic, Google, Meta, and xAI. The model choice can be made per agent or per use case, giving architects flexibility to match model capabilities and costs to specific requirements.</p>



<h3 class="wp-block-heading">How does Oracle AI Agent Studio compare to Microsoft Copilot Studio and Salesforce Agentforce?</h3>



<p class="wp-block-paragraph">Oracle&#8217;s primary differentiator is industry-specific depth. While Copilot Studio and Agentforce are strong horizontal platforms tied to their respective ecosystems (Microsoft 365/Dynamics and Salesforce CRM), Oracle ships prebuilt agents and templates for eight industry verticals. Oracle also offers broader model choice (five model providers versus Microsoft&#8217;s OpenAI-primary and Salesforce&#8217;s Einstein-primary approaches).</p>



<h3 class="wp-block-heading">When will Oracle AI Agent Studio for Industry Applications be available?</h3>



<p class="wp-block-paragraph">Oracle has announced planned availability within the next 12 months, initially in the United States, with expansion to all supported OCI Enterprise AI regions over time. The Fusion version of the studio is already generally available at no additional cost to eligible customers.</p>



<h2 class="wp-block-heading">Key Takeaways</h2>



<ul class="wp-block-list">
<li><strong>Industry-first, not horizontal-first:</strong> Oracle is the first major enterprise vendor to ship an agent platform with prebuilt agents and templates for eight specific industry verticals — a strategic bet on domain depth over platform breadth.</li>



<li><strong>MCP as the connectivity standard:</strong> By adopting Model Context Protocol for tool connectivity, Oracle aligns with the open standard gaining traction across the enterprise agent ecosystem, including Anthropic&#8217;s Claude, Appian, and the broader open-source community.</li>



<li><strong>Multi-model flexibility:</strong> Support for OpenAI, Anthropic, Google, Meta, and xAI models through OCI Enterprise AI avoids single-vendor lock-in at the model layer — a significant advantage as model capabilities and pricing shift quarterly.</li>



<li><strong>Cross-application orchestration:</strong> Multi-agent teams can coordinate workflows spanning both Industry and Fusion Cloud Applications without data movement or custom integration — the closest thing to &#8220;agents that work where the data lives&#8221; in the current market.</li>



<li><strong>19 months of iteration:</strong> This is not a v1 launch. Oracle has iterated from 50 agents to 1,000+ since September 2024, with the Agent Marketplace, Agentic Applications Builder, and AI-native builder releases building on each other. The Industry Applications extension adds vertical depth to a platform that already has horizontal breadth.</li>



<li><strong>Governance built in:</strong> Role-based guardrails, audit trails, and AI-unit consumption monitoring reflect Oracle&#8217;s enterprise DNA — agents cannot see data the user cannot see, and every action is logged.</li>



<li><strong>Platform-native limitation:</strong> The studio&#8217;s value proposition is strongest for organizations already running Oracle Industry and Fusion applications. Non-Oracle shops should evaluate competing platforms or cross-platform orchestration solutions.</li>
</ul>



<h2 class="wp-block-heading">External References</h2>



<ol class="wp-block-list">
<li>Oracle Press Release, &#8220;Oracle AI Agent Studio for Industry Applications Helps Accelerate Business Outcomes,&#8221; October 21, 2026 — <a href="https://www.oracle.com/news/announcement/oracle-ai-agent-studio-for-industry-applications-helps-accelerate-business-outcomes-2026-10-21/" target="_blank" rel="noopener nofollow">oracle.com</a></li>



<li>Unite.AI, &#8220;Oracle Debuts Low-Code Studio for Building Industry-Specific AI Agents,&#8221; October 2026 — <a href="https://www.unite.ai/oracle-debuts-low-code-studio-for-building-industry-specific-ai-agents/" target="_blank" rel="noopener nofollow">unite.ai</a></li>



<li>WebProNews, &#8220;Oracle Brings AI Agent Studio to Its Industry Applications,&#8221; October 2026 — <a href="https://www.webpronews.com/oracle-ai-agent-studio-industry-applications" target="_blank" rel="noopener nofollow">webpronews.com</a></li>



<li>Cloud Wars, &#8220;Steve Miranda Outlines Oracle&#8217;s Agentic AI Vision and First Wave of Agentic Applications,&#8221; April 2026 — <a href="https://cloudwars.com/innovation-leadership/steve-miranda-outlines-oracles-agentic-ai-vision-and-first-wave-of-agentic-applications/" target="_blank" rel="noopener nofollow">cloudwars.com</a></li>



<li>Argano, &#8220;Oracle Has Become One of the Most Aggressive AI Innovators in Enterprise Tech,&#8221; 2026 — <a href="https://argano.com/insights/articles/oracle-has-become-one-of-the-most-aggressive-ai-innovators-in-enterprise-tech.html" target="_blank" rel="noopener nofollow">argano.com</a></li>



<li>CryptoBriefing, &#8220;Oracle Launches AI Agent Studio for Industry Applications,&#8221; October 2026 — <a href="https://cryptobriefing.com/oracle-ai-agent-studio-industry-applications/" target="_blank" rel="noopener nofollow">cryptobriefing.com</a></li>



<li>Oracle, &#8220;AI Agents to Help Automate Dozens of Processes,&#8221; 2024 — <a href="https://www.oracle.com/applications/fusion-ai/steve-miranda-ai-agents-business-productivity/" target="_blank" rel="noopener nofollow">oracle.com</a></li>



<li>PRNewswire, Official Press Release, October 21, 2026 — <a href="https://www.prnewswire.com/news-releases/oracle-ai-agent-studio-for-industry-applications-helps-accelerate-business-outcomes-302914801.html" target="_blank" rel="noopener nofollow">prnewswire.com</a></li>
</ol>



<p class="wp-block-paragraph"></p>
]]></content:encoded>
					
					<wfw:commentRss>https://rpabotsworld.com/oracle-ai-agent-studio-industry-applications-guide/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<media:thumbnail url="https://rpabotsworld.com/wp-content/uploads/2026/10/Oracle-AI-Agent-Studio-for-Industry-Applications-—-Complete-Guide.png" />	</item>
		<item>
		<title>Appian Agentic Automation Platform: MCP Integration, Data Fabric, and Gartner #1 Rankings — The Complete Guide for Agentic AI Architects (2026)</title>
		<link>https://rpabotsworld.com/appian-agentic-automation-mcp-data-fabric-gartner-guide/</link>
					<comments>https://rpabotsworld.com/appian-agentic-automation-mcp-data-fabric-gartner-guide/#respond</comments>
		
		<dc:creator><![CDATA[Satish Prasad]]></dc:creator>
		<pubDate>Wed, 14 Oct 2026 00:20:59 +0000</pubDate>
				<category><![CDATA[RPA & Bot Automation]]></category>
		<category><![CDATA[MCP]]></category>
		<guid isPermaLink="false">https://rpabotsworld.com/?p=32442</guid>

					<description><![CDATA[Appian agentic automation explained — MCP integration, Data Fabric, Snowflake partnership, and why Gartner ranked it #1 in two BOAT use cases.]]></description>
										<content:encoded><![CDATA[
<p class="wp-block-paragraph">A platform that most agentic AI conversations skip just topped two Gartner use-case rankings, shipped native MCP servers, and partnered with Snowflake to give agents governed access to enterprise data. While UiPath and Automation Anywhere dominate headlines, Appian has been quietly building the orchestration layer that enterprise AI agents actually need: process-centric governance, a unified data fabric, and interoperability with every major agent framework through MCP. This guide breaks down what changed, why it matters, and how it compares to the platforms you already know.</p>



<h2 class="wp-block-heading">Why Appian Deserves Your Attention Now</h2>



<p class="wp-block-paragraph">Appian (NASDAQ: APPN) has operated in the low-code automation and business process management space since 2000, long predating the RPA wave. But in 2026, the company made a strategic pivot that puts it squarely in the agentic AI conversation. At Appian World 2026 in April, the company announced MCP protocol adoption, a Snowflake partnership for data-backed agent decisions, AI-assisted spec-driven development through Appian Composer, and environment-wide AI guardrails — all shipping across releases 26.4 through 26.6.</p>



<p class="wp-block-paragraph">Then in September 2026, the <a href="https://itwire.com/guest-articles/company-news/2026-gartner-critical-capabilities-for-business-orchestration-and-automation-technologies-report-ranks-appian-1-for-the-deterministic-workflow-automation-use-case-and-platform-cons" target="_blank" rel="noopener nofollow">2026 Gartner Critical Capabilities for Business Orchestration and Automation Technologies (BOAT) report</a> ranked Appian #1 for both the Deterministic Workflow Automation use case and the Platform Consolidation use case — out of 20 evaluated vendors. Gartner also named Appian a Leader in the companion Magic Quadrant for BOAT, making it the only Leader ranked in the top three across all five evaluated use cases.</p>



<p class="wp-block-paragraph">This is not a minor positioning shift. Gartner predicts that by 2030, 70% of enterprises will pivot to a consolidated automation platform, up from roughly 10% today. Appian is betting — and Gartner&#8217;s data suggests it is betting correctly — that the winning agentic AI strategy is not building more standalone agents, but embedding agents inside governed business processes with unified data access.</p>



<h2 class="wp-block-heading">Appian&#8217;s Agentic Architecture: How It Actually Works</h2>



<p class="wp-block-paragraph">Appian&#8217;s agentic automation rests on four pillars: interoperability through MCP, a context layer via the Data Fabric, process-centric orchestration, and environment-wide guardrails. Understanding each is essential for architects evaluating the platform against UiPath, Automation Anywhere, or pure-play agent frameworks like LangGraph and CrewAI.</p>



<h3 class="wp-block-heading">Pillar 1: MCP Integration — Bidirectional Agent Interoperability</h3>



<p class="wp-block-paragraph">The <a href="https://siliconangle.com/2026/04/28/appian-adopts-mcp-protocol-partners-snowflake-provide-structure-control-ai-agents/" target="_blank" rel="noopener nofollow">Model Context Protocol (MCP) adoption</a> announced at Appian World 2026 is the most architecturally significant change. Appian&#8217;s implementation is bidirectional:</p>



<p class="wp-block-paragraph"><strong>Outbound (Appian agents calling external tools):</strong> Appian agents can connect to any external MCP server — Snowflake, GitHub, Google Drive, and any other MCP-compatible service — without building custom integrations. Release 26.6 ships a <strong>Generic MCP Connected System</strong> that makes this a configuration step, not a development project. This means an Appian agent orchestrating a claims-processing workflow can federate a Snowflake query, pull a document from Google Drive, and update a GitHub issue — all within a single governed process.</p>



<p class="wp-block-paragraph"><strong>Inbound (external agents calling Appian):</strong> Organizations can administer their own <strong>Appian MCP Server</strong>, which exposes record types, process models, and expression rules as secure, discoverable tools for external agents. This is a critical differentiator: Google Vertex AI agents, LangGraph chains, or any MCP-compatible agent framework can invoke Appian&#8217;s business rules, trigger processes, and read/write data through the Data Fabric — all subject to Appian&#8217;s security and governance controls.</p>



<p class="wp-block-paragraph">For RPA teams already running UiPath or Automation Anywhere alongside Appian, this bidirectional MCP means agents built on one platform can call capabilities on the other without middleware. It is the kind of interoperability that the <a href="https://rpabotsworld.com/a2a-joins-mcp-aaif-agentic-ai-architects-guide/">Agentic AI Foundation&#8217;s vision for MCP and A2A</a> was designed to enable.</p>



<h3 class="wp-block-heading">Pillar 2: Data Fabric — The Context Layer Agents Need</h3>



<p class="wp-block-paragraph">Agents without context make bad decisions. This is the central insight driving Appian&#8217;s Data Fabric, which provides a unified metadata model that maps fragmented enterprise data to shared business terms. When an agent encounters &#8220;Customer&#8221; or &#8220;Order,&#8221; the Data Fabric ensures it resolves to a canonical definition that spans CRM, ERP, and every connected system.</p>



<p class="wp-block-paragraph">Release 26.6 extends the Data Fabric with several agent-critical capabilities:</p>



<ul class="wp-block-list">
<li><strong>Query Data Fabric tool for external agents:</strong> External agents can discover synced record types and retrieve data via SQL, while Appian enforces row-level and field-level security. No agent gets data it should not have, regardless of which framework spawned it.</li>



<li><strong>Snowflake native support:</strong> Record types connect directly to Snowflake data without syncing, with full relationship mapping and record-level security. Combined with the MCP-enabled integration, agents can interact with Snowflake Cortex AI for data-backed decisions within governed workflows.</li>



<li><strong>Combined smart search:</strong> Semantic and lexical search work together in <code>a!queryRecordType()</code>, and semantic search extends to <code>a!relatedRecordData()</code>. Agents can find related records by meaning, not just exact match.</li>



<li><strong>Autoscale for Data Fabric:</strong> The data service scales automatically on compute-optimized infrastructure, delivering up to 2x faster queries for heavy workloads — essential when multiple agents are querying simultaneously.</li>



<li><strong>External document metadata sync:</strong> Document metadata from connected systems syncs with core business data, and smart search indexes PDFs and DOCX files up to 200MB.</li>
</ul>



<p class="wp-block-paragraph">As Baris Gultekin, VP of AI at Snowflake, stated at the partnership announcement: &#8220;Enterprises don&#8217;t need more AI experiments, they need AI that delivers real business outcomes on governed data.&#8221; The Appian-Snowflake integration positions the Data Fabric as the semantic bridge between Snowflake&#8217;s AI Data Cloud and Appian&#8217;s process orchestration — giving agents both the data and the operational context to act on it.</p>



<h3 class="wp-block-heading">Pillar 3: Process-Centric Agent Orchestration</h3>



<p class="wp-block-paragraph">This is where Appian&#8217;s approach diverges most sharply from pure-play agent frameworks. In LangGraph, CrewAI, or AutoGen, agents are typically orchestrated through code — you define agent roles, tool access, and interaction patterns programmatically. Appian inverts this: <strong>agents run inside business processes</strong>. The process model defines each agent&#8217;s goals, responsibilities, accessible data, and how its actions connect to the wider workflow.</p>



<p class="wp-block-paragraph">The practical implications for enterprise teams are significant:</p>



<p class="wp-block-paragraph"><strong>Multi-agent collaboration within process boundaries:</strong> A lead agent can dynamically invoke specialized agents — whether inside Appian or in external systems via MCP — for tasks like triage, data extraction, and approval routing. All of this runs inside a single process boundary, which means decisions, escalations, and exceptions follow governed paths rather than ad-hoc agent-to-agent negotiation.</p>



<p class="wp-block-paragraph"><strong>Process HQ for observability:</strong> Release 26.6 adds process reports built on process analysis results, interactive process diagram insights (click an activity to create a focused KPI with suggested insights), and automation type icons that distinguish whether a human, RPA bot, AI agent, or integration performed each step. For teams already running mixed human-bot-agent workflows, this is the visibility that <a href="https://rpabotsworld.com/forrester-wave-adaptive-process-orchestration-q3-2026-guide/">Forrester&#8217;s Adaptive Process Orchestration framework</a> describes as essential.</p>



<p class="wp-block-paragraph"><strong>Parallel tool execution:</strong> Release 26.6 enables agents to run independent tool calls concurrently, reducing both wall-clock time and per-action costs. For a claims-processing agent that needs to verify coverage, pull patient records, and check provider status, this means those calls happen simultaneously rather than sequentially.</p>



<p class="wp-block-paragraph"><strong>Agent chat integration:</strong> The new <code>a!agentChatField</code> component embeds a conversational interface directly inside application layouts, so users interact with agents in context rather than switching to a separate chatbot.</p>



<h3 class="wp-block-heading">Pillar 4: Environment-Wide AI Guardrails</h3>



<p class="wp-block-paragraph">Most agent platforms apply governance at the individual agent or application level. Appian takes an environment-wide approach: the new <strong>AI Guardrails</strong> tab in the Admin Console screens all AI inputs and outputs — across every agent, every AI Skill, every generative AI call in the environment — for three named risks:</p>



<ul class="wp-block-list">
<li><strong>Prompt injection:</strong> Detects and blocks attempts to manipulate agent behavior through crafted inputs</li>



<li><strong>PII leakage:</strong> Prevents personally identifiable information from flowing to unauthorized AI models or outputs</li>



<li><strong>Toxic content:</strong> Filters harmful or inappropriate content from AI-generated responses</li>
</ul>



<p class="wp-block-paragraph">Administrators configure these policies once, and they apply to every AI interaction in the Appian environment. Combined with the Data Fabric&#8217;s row-level and field-level access controls, this creates a governance model where agents are constrained by both data security and content safety — without requiring each development team to implement their own guardrails.</p>



<p class="wp-block-paragraph">Release 26.6 also adds <strong>model family controls</strong> (administrators can disable specific model families to meet regulatory requirements), <strong>LLM gateway support</strong> (route all AI requests through an internal gateway), and support for <strong>Azure OpenAI</strong>, <strong>Amazon Nova 2 Lite</strong>, and <strong>Claude Sonnet 4.6</strong> — giving teams flexibility in model selection while maintaining centralized governance.</p>



<h2 class="wp-block-heading">Appian Composer: AI-Assisted Spec-Driven Development</h2>



<p class="wp-block-paragraph">Appian Composer represents a fundamentally different approach to application modernization. Rather than building applications from scratch or manually translating legacy requirements, Composer uses AI to extract specifications from existing legacy applications — including their user interfaces, data models, and process flows — and generates a visual modernization plan.</p>



<p class="wp-block-paragraph">The workflow looks like this:</p>



<ol class="wp-block-list">
<li><strong>Input legacy artifacts:</strong> Upload Excel files (Composer interprets formulas, conditional logic, and macros), XML, YAML, PPTX files, BPMN diagrams, and up to 20 images of legacy UI screenshots</li>



<li><strong>AI extracts specifications:</strong> The system analyzes inputs and produces a structured plan covering the new application&#8217;s interfaces, data entities, business logic, and process flows</li>



<li><strong>Iterative refinement:</strong> Use a clickable app preview to review the application flow during the planning phase, before any design objects exist</li>



<li><strong>AI developer agents build:</strong> Human-supervised AI agents complete development tasks against the approved specifications, with a kanban Board tab for tracking and assigning tasks to Composer</li>
</ol>



<p class="wp-block-paragraph">CTO and founder Mike Beckley describes Composer as built on &#8220;Appian&#8217;s new open MCP — a model-driven representation of your complete application estate, including requirements, apps, data entities, logic, workflows, security and governance rules, integrations, and dependencies.&#8221; This estate representation is exposed as context for both human developers and AI agents.</p>



<p class="wp-block-paragraph">Critically, Appian&#8217;s developer MCP servers also support external AI development tools. Teams using <strong>Anthropic&#8217;s Claude Code</strong> or <strong>Amazon&#8217;s Kiro</strong> can build and update Appian applications through those tools, using MCP as the integration layer. This is a pragmatic move: rather than forcing teams to abandon their preferred development tools, Appian makes its entire platform accessible through the same protocol that <a href="https://rpabotsworld.com/microsoft-agent-framework-1-0-foundry-agentic-ai-architect-guide/">Microsoft Agent Framework</a> and other major platforms support.</p>



<h2 class="wp-block-heading">RPA Capabilities: What&#8217;s New in 26.4–26.6</h2>



<p class="wp-block-paragraph">Appian&#8217;s RPA module, while less headline-grabbing than its agentic features, received meaningful updates across the 26.4–26.6 release cycle that matter for teams running production bots:</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Release</th><th>Feature</th><th>Impact</th></tr></thead><tbody><tr><td>26.6</td><td>Native close-tab action + improved Get Table Value</td><td>Handles non-standard web tables and tab management without custom scripts</td></tr><tr><td>26.5</td><td>Reserve Credential by UUID</td><td>Fixed input for credential reservation — critical for multi-robot environments</td></tr><tr><td>26.5</td><td>Concurrent robot authentication</td><td>Eliminates queue instability when many robots connect simultaneously</td></tr><tr><td>26.5</td><td>Faster HA support files</td><td>Upload, download, rename, move, and extract operations complete in seconds in HA environments</td></tr><tr><td>26.4</td><td>High availability for RPA</td><td>Available on request — removes the single-point-of-failure risk for production RPA</td></tr><tr><td>26.4</td><td>Credential and VDI REST APIs</td><td>Automate password rotation and autologin updates without the RPA Console</td></tr></tbody></table></figure>



<p class="wp-block-paragraph">One notable deprecation: <strong>RPA Queues and Schedules were removed in RPA 9.22</strong>. Teams must migrate schedules to Timer Events and replace queues with Low-Code Process (LCP) patterns. This is consistent with Appian&#8217;s broader strategy of pulling RPA into the process orchestration layer rather than treating it as a standalone capability.</p>



<h2 class="wp-block-heading">How Appian Compares: The Enterprise Automation Decision Matrix</h2>



<p class="wp-block-paragraph">For teams evaluating Appian against the dominant RPA and automation platforms, here is how the key capabilities stack up in the agentic era:</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Capability</th><th>Appian</th><th>UiPath</th><th>Automation Anywhere</th><th>SS&amp;C Blue Prism</th></tr></thead><tbody><tr><td><strong>MCP Support</strong></td><td>Native bidirectional (26.6)</td><td>Not announced</td><td>Not announced</td><td>Not announced</td></tr><tr><td><strong>Agent Orchestration Model</strong></td><td>Process-centric (agents inside processes)</td><td>Autopilot + Agent Builder</td><td>AI Agent Studio</td><td>WorkHQ intelligent orchestration</td></tr><tr><td><strong>Unified Data Layer</strong></td><td>Data Fabric with semantic search</td><td>Data Service + Integration Service</td><td>CoE Manager data pipeline</td><td>WorkHQ data connections</td></tr><tr><td><strong>AI Guardrails</strong></td><td>Environment-wide (prompt injection, PII, toxicity)</td><td>AI Trust Layer</td><td>Governance controls</td><td>Platform-level security</td></tr><tr><td><strong>Low-Code Process Design</strong></td><td>Core strength — 25+ years</td><td>StudioX (citizen), Studio (pro)</td><td>Process Editor</td><td>Blue Prism Studio</td></tr><tr><td><strong>Process Mining</strong></td><td>Process HQ (native)</td><td>Process Mining (acquired)</td><td>Process Discovery</td><td>Limited native</td></tr><tr><td><strong>Spec-Driven Modernization</strong></td><td>Composer (AI extracts from legacy apps)</td><td>Autopilot for developers</td><td>Not available</td><td>Not available</td></tr><tr><td><strong>Gartner BOAT 2026 Ranking</strong></td><td>#1 in 2 use cases; Leader in MQ</td><td>Leader in MQ</td><td>Leader in MQ</td><td>Positioned in MQ</td></tr><tr><td><strong>Core Pricing Model</strong></td><td>Per-user + platform fees</td><td>Per-robot + AI Units</td><td>Per-bot-runner</td><td>Per-digital-worker</td></tr></tbody></table></figure>



<p class="wp-block-paragraph"><strong>Use Appian when:</strong> Your primary need is orchestrating complex, multi-system business processes where agents, humans, and RPA bots collaborate inside governed workflows. Appian excels when process design is the foundation and AI/RPA are capabilities embedded within it — particularly for organizations consolidating from multiple automation tools (Gartner&#8217;s #1 Platform Consolidation ranking reflects this). The MCP integration makes it especially strong when you need agents from multiple frameworks (Google Vertex AI, LangGraph, Claude Code) to interact with a central process orchestration layer.</p>



<p class="wp-block-paragraph"><strong>Use UiPath when:</strong> Your primary need is large-scale attended and unattended RPA with an expanding AI layer. UiPath&#8217;s robot ecosystem is deeper, its marketplace larger, and its <a href="https://rpabotsworld.com/automation-anywhere-vs-uipath-agentic-procure-to-pay-2026/">agentic capabilities for specific processes like P2P</a> are more mature in production deployments. The Everest Group Process Orchestration PEAK Matrix 2026 named UiPath a Leader and Star Performer, confirming its strength in orchestration as well.</p>



<p class="wp-block-paragraph"><strong>Use Automation Anywhere when:</strong> Your primary need is cloud-native RPA with embedded AI skills, particularly for document processing and data extraction at scale. AA&#8217;s AI Agent Studio offers a more code-first agent development experience compared to Appian&#8217;s process-first model.</p>



<p class="wp-block-paragraph"><strong>Use Blue Prism when:</strong> Your primary need is enterprise-grade RPA in regulated industries (banking, insurance, healthcare) where <a href="https://rpabotsworld.com/ssc-blue-prism-workhq-agentic-automation-platform-guide/">WorkHQ&#8217;s intelligent orchestration</a> and audit trail depth matter more than agentic flexibility. Note: the Blue Prism WorkHQ draft at rpabotsworld.com covers this in depth.</p>



<h2 class="wp-block-heading">The Snowflake Partnership: What It Means for Data-Driven Agents</h2>



<p class="wp-block-paragraph">The Appian-Snowflake partnership is more than a press release. The direct MCP-enabled integration between Appian&#8217;s Data Fabric and Snowflake&#8217;s AI Data Cloud creates a technical architecture where:</p>



<ol class="wp-block-list">
<li><strong>Process orchestration meets data warehouse:</strong> Appian agents can query Snowflake data through the Data Fabric without syncing it locally, maintaining Snowflake&#8217;s security model while operating within Appian&#8217;s process governance</li>



<li><strong>Cortex AI access within workflows:</strong> Agents interact with Snowflake Cortex AI — Snowflake&#8217;s serverless machine learning service — for real-time predictions and classifications during process execution</li>



<li><strong>Unified metadata model:</strong> The Data Fabric&#8217;s metadata layer extends across Snowflake tables, giving agents semantic context about how data is structured and connected — the shared business vocabulary that prevents agents from misinterpreting fields across systems</li>
</ol>



<p class="wp-block-paragraph">For organizations already invested in Snowflake as their data platform, this partnership means the path from data insight to automated action passes through a governed process layer rather than an ad-hoc integration. The practical impact: a supply chain agent that detects an anomaly in Snowflake&#8217;s data can trigger an Appian workflow that notifies procurement, creates a purchase order, and routes it for approval — all within a single governed flow, with every step auditable.</p>



<h2 class="wp-block-heading">Real-World Deployment: Global Excel Management</h2>



<p class="wp-block-paragraph">Global Excel Management, a healthcare risk management provider, offers the most publicly referenced deployment case. The company is using Appian to unify claims processing workflows — from intake through adjudication — that were previously fragmented across multiple systems and manual handoffs.</p>



<p class="wp-block-paragraph">Pascal Tanguay, SVP of Global Technology Services at Global Excel Management, stated: &#8220;With Appian, our processes will be unified.&#8221; The implementation uses AI to reduce redundant tasks while maintaining the compliance requirements inherent in healthcare claims processing. Appian claims its AI processes documents with over 95% accuracy, approximately 35% better than legacy alternatives — though this figure comes from Appian&#8217;s own benchmarks rather than an independent audit.</p>



<p class="wp-block-paragraph">For RPA teams in healthcare, insurance, and financial services — industries where process compliance is non-negotiable — the Global Excel case illustrates Appian&#8217;s core value proposition: you can add AI agents to your workflows without losing the governance and auditability that regulated industries require.</p>



<h2 class="wp-block-heading">What&#8217;s Coming Next: The Roadmap Signals</h2>



<p class="wp-block-paragraph">Appian has signaled several capabilities as &#8220;coming soon&#8221; without specific release dates:</p>



<ul class="wp-block-list">
<li><strong>Objective optimization guidance:</strong> Users will be able to give AI agents guidance on which business objectives to optimize and receive recommended improvements that can be applied safely</li>



<li><strong>Agent memory across processes:</strong> The ability to track agent performance and apply an agent&#8217;s memory across different processes to improve decisions over time — a feature that moves toward the persistent, learning agents that <a href="https://rpabotsworld.com/openai-devday-2026-dots-agents-gpt-6-1-sol-guide/">OpenAI&#8217;s Dots represent</a> in the consumer space</li>



<li><strong>Work IQ integration signals:</strong> While not Appian-specific, Microsoft&#8217;s Work IQ preview (October 2026) grounds Copilot agents in live Dynamics 365 and Power Platform data — and Appian&#8217;s MCP servers could potentially expose Appian processes and data to Work IQ agents, creating another integration pathway</li>
</ul>



<p class="wp-block-paragraph">These roadmap items suggest Appian is moving toward adaptive agents that learn from process execution — a significant step beyond the deterministic workflow automation where Gartner currently ranks them #1.</p>



<h2 class="wp-block-heading">Implementation Considerations for Architects</h2>



<p class="wp-block-paragraph">Before deploying Appian&#8217;s agentic capabilities, architects should evaluate these practical factors:</p>



<p class="wp-block-paragraph"><strong>Licensing and cost structure:</strong> Appian uses a per-user plus platform fee model. AI agent actions consume &#8220;AI Actions&#8221; that are tracked and attributed to Appian groups for cost management (release 26.6 adds Usage Groups input and AI Actions Consumed output). Unlike UiPath&#8217;s AI Units or Automation Anywhere&#8217;s per-bot pricing, Appian&#8217;s consumption model ties costs to business process execution rather than robot count — which can be cheaper for process-heavy, bot-light scenarios and more expensive for high-volume, low-process RPA.</p>



<p class="wp-block-paragraph"><strong>Model flexibility:</strong> As of 26.6, Appian supports Azure OpenAI, Amazon Nova 2 Lite, Claude Sonnet 4.6, and custom LLM gateway endpoints. The model family controls let administrators disable specific model families for regulatory compliance — essential for teams in healthcare (HIPAA) or finance (SOX/PCI-DSS) where data residency matters.</p>



<p class="wp-block-paragraph"><strong>Migration path:</strong> For teams running UiPath or Automation Anywhere bots, Appian is not a rip-and-replace proposition. The MCP integration means external bots can call Appian processes and data, and Appian agents can trigger external RPA bots. The practical migration strategy is to use Appian as the orchestration layer while keeping existing RPA investments operational — precisely the platform consolidation scenario Gartner validated.</p>



<p class="wp-block-paragraph"><strong>Developer experience:</strong> Appian Composer and the developer MCP servers significantly lower the barrier for AI-native developers. But Appian&#8217;s visual process designer remains the primary development surface, which means teams accustomed to code-first platforms (LangGraph, CrewAI) will need to adapt to a low-code workflow. The Claude Code and Kiro integrations mitigate this by letting developers use familiar tools to build Appian components.</p>



<h2 class="wp-block-heading">Key Takeaways</h2>



<ul class="wp-block-list">
<li><strong>Appian ranked #1 in two Gartner BOAT Critical Capabilities use cases</strong> (Deterministic Workflow Automation and Platform Consolidation) and is a Leader in the Magic Quadrant — positioning it as the top choice for enterprises consolidating automation platforms</li>



<li><strong>Bidirectional MCP integration</strong> makes Appian one of the first enterprise automation platforms to support native interoperability with the broader agent ecosystem — agents built on Google Vertex AI, LangGraph, or Claude Code can call Appian processes, and vice versa</li>



<li><strong>The Snowflake partnership</strong> combines process orchestration with AI Data Cloud access, enabling data-backed agent decisions within governed workflows — critical for enterprises with Snowflake as their data platform</li>



<li><strong>Environment-wide AI guardrails</strong> apply to every AI interaction across the platform, screening for prompt injection, PII leakage, and toxic content — a more comprehensive approach than per-agent governance</li>



<li><strong>Appian Composer&#8217;s spec-driven development</strong> extracts specifications from legacy applications and uses AI developer agents to modernize them under human supervision — a practical path for the massive legacy-to-agentic migration wave</li>



<li><strong>Process-centric orchestration</strong> is the fundamental differentiator: agents run inside business processes rather than operating independently, which means governance, escalation, and audit trails are built into the architecture rather than bolted on</li>



<li><strong>RPA capabilities continue to mature</strong> with high availability, concurrent authentication, and REST APIs for credential management — but the strategic direction is clear: RPA is becoming one capability within a broader orchestration platform</li>
</ul>



<h2 class="wp-block-heading">Frequently Asked Questions</h2>



<h3 class="wp-block-heading">Does Appian replace UiPath or Automation Anywhere?</h3>



<p class="wp-block-paragraph">Not necessarily. Appian&#8217;s MCP integration means it can orchestrate existing UiPath or Automation Anywhere bots as part of broader business processes. The platform consolidation scenario — where Appian becomes the orchestration layer on top of existing RPA investments — is the approach Gartner validated by ranking Appian #1 for that use case. For greenfield automation projects, Appian&#8217;s native RPA plus agentic capabilities may eliminate the need for a separate RPA platform.</p>



<h3 class="wp-block-heading">How does Appian&#8217;s MCP integration compare to UiPath&#8217;s agent capabilities?</h3>



<p class="wp-block-paragraph">UiPath&#8217;s Autopilot and Agent Builder provide agent capabilities within the UiPath ecosystem. Appian&#8217;s bidirectional MCP integration is broader: any MCP-compatible agent framework (LangGraph, Google Vertex AI, Claude Code, and others) can both call Appian processes and be called by Appian agents. This interoperability advantage matters most for enterprises running heterogeneous AI stacks rather than a single vendor&#8217;s ecosystem.</p>



<h3 class="wp-block-heading">What AI models does Appian support for its agents?</h3>



<p class="wp-block-paragraph">As of release 26.6, Appian supports Azure OpenAI, Amazon Nova 2 Lite, and Anthropic Claude Sonnet 4.6 natively, with a custom LLM gateway option for routing to any model endpoint. Administrators can enable or disable specific model families per environment for regulatory compliance.</p>



<h3 class="wp-block-heading">Is Appian suitable for pure RPA use cases without AI agents?</h3>



<p class="wp-block-paragraph">Yes. Appian&#8217;s RPA module operates independently of its agentic features and received meaningful updates in 26.4–26.6 (high availability, concurrent authentication, credential REST APIs). However, the strategic direction is toward unified orchestration — teams choosing Appian for RPA alone may find UiPath or Automation Anywhere more focused on that specific capability.</p>



<h3 class="wp-block-heading">What does Appian&#8217;s Gartner #1 ranking actually mean in practice?</h3>



<p class="wp-block-paragraph">Gartner evaluated 20 vendors across five use cases in its Critical Capabilities for BOAT report. Appian scored highest in Deterministic Workflow Automation (structured, rules-based processes) and Platform Consolidation (replacing multiple automation tools with one). It was also the highest-scoring vendor for the Platform Operations and Governance critical capability. This does not mean Appian is the best choice for every automation scenario — but for enterprises consolidating tools and governing complex workflows, Gartner&#8217;s data puts Appian at the top.</p>



<h2 class="wp-block-heading">References</h2>



<ol class="wp-block-list">
<li>Appian Press Release, &#8220;Appian Advances AI in Process to Deliver Enterprise Outcomes at Scale,&#8221; April 28, 2026 — <a href="https://appian.com/about/explore/press-releases/2026/appian-advances-ai-in-process-to-deliver-enterprise-outcomes-at-scale" target="_blank" rel="noopener nofollow">appian.com</a></li>



<li>SiliconANGLE, &#8220;Appian adopts MCP protocol, partners with Snowflake to provide structure and control for AI agents,&#8221; April 28, 2026 — <a href="https://siliconangle.com/2026/04/28/appian-adopts-mcp-protocol-partners-snowflake-provide-structure-control-ai-agents/" target="_blank" rel="noopener nofollow">siliconangle.com</a></li>



<li>iTWire, &#8220;2026 Gartner Critical Capabilities for BOAT Report Ranks Appian #1,&#8221; September 18, 2026 — <a href="https://itwire.com/guest-articles/company-news/2026-gartner-critical-capabilities-for-business-orchestration-and-automation-technologies-report-ranks-appian-1-for-the-deterministic-workflow-automation-use-case-and-platform-cons" target="_blank" rel="noopener nofollow">itwire.com</a></li>



<li>Appian Blog, &#8220;Autonomy Requires Control: How to Orchestrate Enterprise AI Agents,&#8221; 2026 — <a href="https://appian.com/blog/2026/enterprise-AI-agents" target="_blank" rel="noopener nofollow">appian.com</a></li>



<li>Appian Documentation, &#8220;Release Notes 26.4–26.6,&#8221; 2026 — <a href="https://docs.appian.com/suite/help/26.6/Appian_Release_Notes.html" target="_blank" rel="noopener nofollow">docs.appian.com</a></li>



<li>Pulse2, &#8220;Appian Unveils Agentic Automation and AI-Assisted Development Capabilities,&#8221; 2026 — <a href="https://pulse2.com/appian-unveils-agentic-automation-and-ai-assisted-development-capabilities-for-enterprise-process-management/" target="_blank" rel="noopener nofollow">pulse2.com</a></li>



<li>IT Brief Asia, &#8220;Appian adds AI-assisted development &amp; MCP integration,&#8221; 2026 — <a href="https://itbrief.asia/story/appian-adds-ai-assisted-development-mcp-integration" target="_blank" rel="noopener nofollow">itbrief.asia</a></li>



<li>Newgen Software, &#8220;Recognized in Forrester&#8217;s The Adaptive Process Orchestration Software Landscape, Q2 2026&#8221; — <a href="https://newgensoft.com/resources/analyst-report-newgen-recognized-in-forresters-the-adaptive-process-orchestration-software-landscape-q2-2026/" target="_blank" rel="noopener nofollow">newgensoft.com</a></li>
</ol>



<p class="wp-block-paragraph"><em>Disclosure: RPABOTS.WORLD is an independent publication. This article is not sponsored by Appian. All claims are attributed to their sources. Gartner does not endorse any vendor, product, or service depicted in its research publications.</em></p>



<p class="wp-block-paragraph"></p>
]]></content:encoded>
					
					<wfw:commentRss>https://rpabotsworld.com/appian-agentic-automation-mcp-data-fabric-gartner-guide/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>Pega Agentic Process Fabric: The Workflow-First Enterprise Agent Platform Guide (2026)</title>
		<link>https://rpabotsworld.com/pega-agentic-process-fabric-enterprise-agent-guide/</link>
					<comments>https://rpabotsworld.com/pega-agentic-process-fabric-enterprise-agent-guide/#comments</comments>
		
		<dc:creator><![CDATA[Satish Prasad]]></dc:creator>
		<pubDate>Wed, 07 Oct 2026 02:35:32 +0000</pubDate>
				<category><![CDATA[RPA & Bot Automation]]></category>
		<category><![CDATA[A2A Protocol]]></category>
		<category><![CDATA[MCP]]></category>
		<guid isPermaLink="false">https://rpabotsworld.com/?p=32413</guid>

					<description><![CDATA[How Pega Agentic Process Fabric puts workflows before prompts. Architecture, MCP/A2A support, Infinity 26 flat-fee pricing, and what it means for RPA teams.]]></description>
										<content:encoded><![CDATA[
<p class="wp-block-paragraph">Gartner projects that more than 40 percent of agentic AI projects will be canceled by the end of 2027, citing escalating costs and inadequate controls as the primary culprits. Pegasystems looked at that forecast and bet the company&#8217;s next platform release on a single contrarian idea: <strong>don&#8217;t let agents improvise at runtime</strong>. The result is Pega Agentic Process Fabric — an orchestration layer that treats AI agents not as autonomous decision-makers but as participants in governed, auditable workflows. If your enterprise already runs mission-critical processes on Pega, or you are evaluating where agent orchestration fits inside a regulated stack, this guide breaks down the architecture, the pricing model, and the competitive tradeoffs you need to understand before your next planning cycle.</p>



<h2 class="wp-block-heading">What Is Pega Agentic Process Fabric?</h2>



<p class="wp-block-paragraph">Pega Agentic Process Fabric is an extension of Pegasystems&#8217; existing Process Fabric architecture, redesigned to orchestrate AI agents alongside traditional applications, systems, and data sources. Announced at PegaWorld in June 2025 and shipped as part of <strong>Pega Infinity 26</strong> in July 2026, the fabric creates a unified registry where every discoverable agent, workflow, and data asset across the enterprise is cataloged and made available to an overarching orchestration agent.</p>



<p class="wp-block-paragraph">In practical terms, the fabric does three things. First, it <strong>discovers and registers</strong> every Pega workflow, third-party AI agent, system integration, and data source connected to it. Second, it <strong>selects the optimal execution path</strong> — which might be a Pega workflow, an external agent, or a combination — for any given request. Third, it <strong>governs the entire interaction</strong> with step-by-step monitoring, audit trails, and compliance controls that satisfy regulated industries.</p>



<p class="wp-block-paragraph">Pega CEO Alan Trefler framed the positioning bluntly: &#8220;Enterprises want the promise and power of automation that agents could offer, but we don&#8217;t think they want thousands of agents running unchecked, producing unreliable, undesirable results.&#8221; That philosophy shapes every design decision in the platform.</p>



<figure class="wp-block-image size-large"><img decoding="async" width="1200" height="800" src="https://rpabotsworld.com/wp-content/uploads/2026/10/pega-agentic-process-fabric-enterprise-agent-guide-architecture-diagram.svg" alt="Pega Agentic Process Fabric three-layer architecture diagram showing Blueprint, Predictable AI Agents, and orchestration fabric" class="wp-image-32414" title="Pega Agentic Process Fabric: The Workflow-First Enterprise Agent Platform Guide (2026) 2"></figure>



<h2 class="wp-block-heading">The Workflow-First Philosophy: Why Pega Rejects Prompt Engineering</h2>



<p class="wp-block-paragraph">Most enterprise agent platforms let AI models reason their way through tasks at runtime — the agent receives a goal, consults tools and data, and figures out the steps on the fly. Pega takes the opposite approach. AI reasoning happens at <strong>design time</strong>, when Blueprint AI agents help architects create and refine workflows. At runtime, those workflows execute through <strong>semantic AI</strong> — deterministic, auditable, and repeatable.</p>



<p class="wp-block-paragraph">Trefler has been direct about why: &#8220;Anybody who&#8217;s asking people to write prompts where those prompts are going to run their business is setting themselves up for slews of challenges.&#8221; CTO Don Schuerman elaborated on the dual approach: &#8220;What we&#8217;ve taken is an approach that combines the power of agents to reason and think and act with the predictability that you get from workflows, from knowing that there are certain things you want to do in the business where you actually want to follow a predetermined path.&#8221;</p>



<p class="wp-block-paragraph">This is not academic. In financial services, healthcare, and government — industries where Pega has deep installed bases — a single unpredictable agent action in a claims adjudication or loan underwriting workflow could trigger a compliance violation. The workflow-first model means that every agent action maps back to an approved process definition, and every deviation is flagged rather than silently executed.</p>



<h3 class="wp-block-heading">What &#8220;Design-Time Reasoning, Runtime Execution&#8221; Actually Means</h3>



<p class="wp-block-paragraph">When a Pega customer needs a new business process, Blueprint AI agents analyze requirements, existing documentation, and best practices to <em>propose</em> a workflow. Architects review, modify, and approve it. Once approved, that workflow becomes immutable at runtime — the agent follows the steps, collects data, routes decisions, and escalates exceptions exactly as designed. The creative, potentially unpredictable AI reasoning is confined to the design phase, where human oversight is built into every approval gate.</p>



<p class="wp-block-paragraph">This stands in contrast to platforms where agents re-reason each task execution. Pega&#8217;s bet is that enterprises will prefer predictable costs and auditable outcomes over the flexibility of unconstrained runtime reasoning — especially after the first wave of &#8220;agent sprawl&#8221; projects hits cost overruns and compliance issues.</p>



<h2 class="wp-block-heading">Architecture Deep Dive: The Three-Layer Stack</h2>



<p class="wp-block-paragraph">Pega Agentic Process Fabric is built on three integrated layers, each with a distinct role in the agent lifecycle.</p>



<h3 class="wp-block-heading">Layer 1: Pega Blueprint — Design Agents</h3>



<p class="wp-block-paragraph">Blueprint is Pega&#8217;s AI-powered design environment. It functions as an agent itself — ingesting documentation, organizational best practices, customer data, and Pega&#8217;s own platform knowledge to generate optimized workflow definitions. Key capabilities include:</p>



<ul class="wp-block-list">
<li><strong>Automated workflow generation</strong> from natural-language requirements</li>



<li><strong>Legacy system analysis</strong> — upload videos, screenshots, source code, or database schemas of existing applications, and Blueprint reverse-engineers modernized process definitions</li>



<li><strong>Industry template application</strong> — pre-loaded best practices for financial services, healthcare, insurance, and government workflows</li>



<li><strong>SI integration</strong> — system integrators like TCS and Credera can embed proprietary intellectual property into Blueprint&#8217;s knowledge base</li>
</ul>



<p class="wp-block-paragraph">Blueprint is currently available for free with a Pega user ID at <a href="https://www.pega.com/blueprint" target="_blank" rel="noopener nofollow">pega.com/blueprint</a>, a strategic move that puts AI-assisted design in the hands of prospects before they commit to the full platform.</p>



<h3 class="wp-block-heading">Layer 2: Pega Predictable AI Agents — Workflow Execution</h3>



<p class="wp-block-paragraph">Once Blueprint produces an approved workflow, Pega Predictable AI technology converts it into a deployable agent via the <strong>Pega Agent Experience API</strong>. These agents combine workflow reliability with contextual AI reasoning — they follow the process definition step by step but use AI to handle natural-language interactions, document analysis, and decision support within the guardrails of the approved workflow.</p>



<p class="wp-block-paragraph">Two pre-built agents ship with the platform:</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Agent</th><th>Function</th><th>Channels</th></tr></thead><tbody><tr><td><strong>Agentic Assignment Agent</strong></td><td>Proactively contacts employees or customers when additional data or approvals are needed; routes work items to the right person based on context</td><td>Email, chat, telephony</td></tr><tr><td><strong>Document Agent</strong></td><td>Analyzes complex documents through splitting, categorization, boundary detection, confidence scoring, and triage; includes a conversational interface for PDF and image insights</td><td>In-app, API</td></tr></tbody></table></figure>



<p class="wp-block-paragraph">Organizations can build custom agents on top of this layer using Pega Infinity Studio, the AI-native development environment introduced in Infinity 26.</p>



<h3 class="wp-block-heading">Layer 3: Pega Agentic Process Fabric — Orchestration</h3>



<p class="wp-block-paragraph">The fabric itself is the orchestration brain. It maintains the enterprise-wide registry of agents, workflows, data sources, and system integrations. When a request arrives — whether from a customer chatbot, an internal employee portal, or another agent — the fabric:</p>



<ol class="wp-block-list">
<li><strong>Parses intent</strong> and determines what needs to happen</li>



<li><strong>Traverses the registry</strong> to find available agents and workflows that can fulfill the request</li>



<li><strong>Selects the optimal path</strong> — possibly combining a Pega workflow with a third-party agent (e.g., a Salesforce CRM lookup) in a single orchestrated flow</li>



<li><strong>Monitors execution</strong> step by step with full audit trails</li>



<li><strong>Escalates exceptions</strong> to human operators when confidence thresholds are not met</li>
</ol>



<p class="wp-block-paragraph">Multi-channel delivery is built in. Agents can engage users through chatbots, email, voice, SMS, and embedded in-app experiences without rebuilding the underlying workflow for each channel.</p>



<h2 class="wp-block-heading">MCP and A2A Protocol Support: The Interoperability Play</h2>



<p class="wp-block-paragraph">Pega made a strategically significant move by supporting both the <strong>Model Context Protocol (MCP)</strong> and <strong>Agent-to-Agent (A2A)</strong> communication standards. With MCP support, every Pega application becomes an MCP server — meaning third-party agents from Anthropic Claude, Google Gemini, OpenAI, AWS AgentCore, and other MCP-compatible platforms can discover and invoke Pega workflows without custom integration code.</p>



<p class="wp-block-paragraph">The A2A protocol enables direct agent-to-agent orchestration across platforms. In a practical scenario, a customer-facing Claude agent could discover a Pega claims-processing workflow via MCP, invoke it, and receive structured results — all governed by Pega&#8217;s compliance layer. For architects already building on the <a href="https://rpabotsworld.com/a2a-joins-mcp-aaif-agentic-ai-architects-guide/">A2A and MCP standards that joined the Agentic AI Foundation</a>, Pega&#8217;s adoption means one more enterprise-grade platform in the interoperable agent ecosystem.</p>



<p class="wp-block-paragraph">The MCP integration extends to development tools as well. Pega Infinity 26 connects to Claude Code, GitHub Copilot, OpenAI Codex, Claude Cowork, ChatGPT, and Office 365 Copilot through MCP, allowing developers to build and modify Pega workflows using their preferred AI-assisted coding environments.</p>



<h2 class="wp-block-heading">Pega Infinity 26: Flat-Fee AI Pricing Changes the Economics</h2>



<p class="wp-block-paragraph">Released on July 14, 2026, Pega Infinity 26 bundles Agentic Process Fabric with a pricing model that directly targets the biggest objection enterprises have to agentic AI: <strong>unpredictable token costs</strong>.</p>



<p class="wp-block-paragraph">Instead of charging per token — where a complex agent interaction might consume thousands of tokens across multiple LLM calls, making cost forecasting nearly impossible — Pega charges a <strong>flat fee per resolved case</strong>. The company calls this &#8220;Pega Predictable AI&#8221; pricing, and it includes an <strong>AI Token Cost Calculator</strong> so prospects can compare their projected costs against token-based alternatives from competing platforms.</p>



<p class="wp-block-paragraph">The economic argument is straightforward: if an agent resolves a customer service case, a claims triage, or a document classification task, the cost is the same regardless of how many LLM calls the workflow required internally. This eliminates the &#8220;AI token tax&#8221; that has caused some enterprises to throttle agent deployments after seeing their first production bills.</p>



<p class="wp-block-paragraph">Key components of Infinity 26 beyond pricing:</p>



<ul class="wp-block-list">
<li><strong>Pega Infinity Studio</strong> — AI-native development environment integrating coding agents with Blueprint best practices</li>



<li><strong>Pega Customer Engagement Studio</strong> — agent-powered workspace for marketers to create personalized 1:1 campaigns at scale</li>



<li><strong>Enhanced MCP integration</strong> — connects Blueprint, Studio, and the runtime platform to external AI coding tools</li>



<li><strong>Cloud deployment</strong> — available through Pega Cloud or managed cloud environments</li>
</ul>



<h2 class="wp-block-heading">Enterprise Governance: The Compliance Architecture</h2>



<p class="wp-block-paragraph">Governance is not a feature bolted onto the Agentic Process Fabric — it is the structural principle. Every agent interaction, whether initiated by a Pega agent or a connected third-party agent, passes through the same governance layer:</p>



<ul class="wp-block-list">
<li><strong>Authentication and access control</strong> — agent access is restricted based on user identity, role, and the specific workflow being invoked</li>



<li><strong>Agent-to-agent interaction regulation</strong> — the fabric controls which agents can communicate with which other agents, preventing unauthorized lateral movement</li>



<li><strong>Step-by-step action monitoring</strong> — every action taken by every agent is logged with timestamps, inputs, outputs, and the workflow step that triggered it</li>



<li><strong>Compliance reporting</strong> — audit trails can be exported for regulatory review, with the ability to replay any agent interaction from start to finish</li>



<li><strong>Human-in-the-loop escalation</strong> — configurable confidence thresholds trigger automatic escalation to human operators for high-stakes decisions</li>
</ul>



<p class="wp-block-paragraph">CTO Don Schuerman has emphasized that this directly addresses the &#8220;black box problem&#8221; that plagues less governed agent platforms. In regulated industries — banking, insurance, healthcare, government — explaining <em>why</em> an agent made a specific decision is not optional. Because every Pega agent action traces back to an approved workflow step, the audit trail is inherently explainable.</p>



<p class="wp-block-paragraph">For teams already evaluating agent governance frameworks, this is worth comparing against <a href="https://rpabotsworld.com/servicenow-ai-control-tower-enterprise-agent-governance-guide/">ServiceNow&#8217;s AI Control Tower approach</a> and <a href="https://rpabotsworld.com/ibm-watsonx-orchestrate-agentic-control-plane-guide/">IBM watsonx Orchestrate&#8217;s control plane model</a> — each takes a meaningfully different architectural approach to the same enterprise trust problem.</p>



<h2 class="wp-block-heading">Legacy Modernization: Blueprint as a Migration Engine</h2>



<p class="wp-block-paragraph">One capability that differentiates Pega from pure-play agent platforms is Blueprint&#8217;s ability to serve as a legacy modernization engine. The workflow is practical: organizations record their legacy application interfaces, upload the recordings (videos, screenshots, source code, database schemas) to Blueprint, and the AI analyzes the functionality to produce modernized workflow definitions.</p>



<p class="wp-block-paragraph">In a banking demo shown at PegaWorld, Blueprint analyzed a credit card management application and identified four distinct workflows to modernize: credit limit adjustments, account maintenance, transaction processing, and billing. Kerim Akgonul, Chief Product Officer, noted: &#8220;What Blueprint will do is go through and essentially what a human would do in months, it will just basically analyze all the content.&#8221;</p>



<p class="wp-block-paragraph">This matters for the RPA community specifically. Many organizations that deployed RPA bots five to eight years ago to automate legacy system interactions now face a choice: maintain the bots as the legacy systems evolve, or modernize the underlying processes. Pega is positioning Blueprint as the path from &#8220;screen-scraping bots on a legacy app&#8221; to &#8220;governed agents on a modern workflow platform&#8221; — a migration pitch aimed squarely at enterprises sitting on aging RPA estates.</p>



<p class="wp-block-paragraph">With 68 percent of IT decision makers reporting that legacy systems prevent modern technology adoption, the modernization play is arguably as strategically important as the agent orchestration capabilities.</p>



<h2 class="wp-block-heading">How Pega Compares: Enterprise Agent Platform Landscape</h2>



<p class="wp-block-paragraph">The enterprise agent orchestration market is forming rapidly, with several major platforms competing for the same budget line item. Here is how Pega&#8217;s approach compares to the leading alternatives:</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Dimension</th><th>Pega Agentic Process Fabric</th><th>Salesforce Agentforce</th><th>ServiceNow AI Control Tower</th><th>IBM watsonx Orchestrate</th></tr></thead><tbody><tr><td><strong>Core philosophy</strong></td><td>Workflow-first: AI reasons at design time, workflows execute at runtime</td><td>Agent-first: pre-built role-based agents with configurable guardrails</td><td>Governance-first: centralized control tower monitors autonomous agents</td><td>Skill-first: agents composed from reusable skill libraries</td></tr><tr><td><strong>MCP support</strong></td><td>Yes — every app becomes an MCP server</td><td>Yes — via Agentforce platform</td><td>Limited</td><td>Yes — via watsonx.ai integration</td></tr><tr><td><strong>A2A protocol</strong></td><td>Yes</td><td>Yes</td><td>Partial</td><td>Yes</td></tr><tr><td><strong>Pricing model</strong></td><td>Flat fee per resolved case</td><td>Per conversation / per agent</td><td>Subscription-based</td><td>Subscription + token consumption</td></tr><tr><td><strong>Legacy modernization</strong></td><td>Built-in via Blueprint AI</td><td>Limited</td><td>Limited</td><td>Via watsonx Code Assistant</td></tr><tr><td><strong>Pre-built agents</strong></td><td>Assignment Agent, Document Agent, Blueprint design agents</td><td>Sales, Service, Marketing, Commerce, and custom agents</td><td>IT, HR, Customer Service agents</td><td>HR, procurement, IT agents</td></tr><tr><td><strong>Best fit</strong></td><td>Regulated industries with existing Pega investments and complex workflow requirements</td><td>Salesforce-centric organizations wanting fast agent deployment</td><td>ITSM-heavy enterprises wanting centralized agent governance</td><td>Multi-cloud enterprises wanting vendor-neutral orchestration</td></tr></tbody></table></figure>



<p class="wp-block-paragraph">The philosophical difference is the key differentiator. Salesforce and ServiceNow let agents reason at runtime within guardrails. Pega confines all reasoning to the design phase. Neither approach is universally superior — the right choice depends on whether your processes are well-defined enough to model as workflows (Pega&#8217;s sweet spot) or require the flexibility of runtime reasoning (where Salesforce and ServiceNow have an edge).</p>



<p class="wp-block-paragraph">For a deeper dive into how <a href="https://rpabotsworld.com/salesforce-agentforce-job-ready-agents-long-horizon-runtime-guide-3/">Salesforce Agentforce&#8217;s job-ready agents</a> compare, or how <a href="https://rpabotsworld.com/microsoft-agent-framework-1-0-foundry-agentic-ai-architect-guide/">Microsoft&#8217;s Agent Framework 1.0</a> fits into the picture, see our dedicated guides.</p>



<h2 class="wp-block-heading">Real-World Use Cases and Customer Signals</h2>



<p class="wp-block-paragraph">While Pega has not published detailed case studies from Agentic Process Fabric deployments (the product only reached general availability in July 2026), several customer signals indicate where early traction is forming:</p>



<h3 class="wp-block-heading">Financial Services</h3>



<p class="wp-block-paragraph">Erica van de Ven, Global Head of Financial Economic Crime Tech at <strong>Rabobank</strong>, stated: &#8220;We see incredible potential for deploying AI agents to improve efficiencies and deliver&#8221; better outcomes in financial crime detection. For banks, the governed workflow model aligns with anti-money laundering (AML) and know-your-customer (KYC) requirements where every decision must be documented and defensible.</p>



<h3 class="wp-block-heading">Insurance Claims Processing</h3>



<p class="wp-block-paragraph">Claims adjudication is a classic Pega stronghold. With Agentic Process Fabric, a claims workflow can now incorporate a Document Agent to analyze submitted evidence, an Assignment Agent to route complex cases to specialist adjusters, and third-party agents (via MCP) to pull data from external systems — all within a single governed flow that satisfies state insurance regulatory requirements.</p>



<h3 class="wp-block-heading">Customer Service Transformation</h3>



<p class="wp-block-paragraph">Pega Customer Engagement Studio, part of Infinity 26, enables contact centers to deploy agents that handle routine inquiries through workflow-governed conversations while escalating complex cases. The flat-fee pricing is particularly attractive here — high-volume customer service operations can budget agent costs per case rather than worrying about token consumption spikes during peak periods.</p>



<h3 class="wp-block-heading">Legacy System Migration</h3>



<p class="wp-block-paragraph">Organizations using Blueprint to modernize legacy applications represent a less obvious but potentially high-value use case. Credera CEO Mahesh Agrawal highlighted the unified directory spanning both Pega and non-Pega systems, enabling &#8220;conversational agents to collaborate, trigger, and complete workflows seamlessly&#8221; — a capability that matters when the modernization target is not replacing the legacy system entirely but wrapping it in governed agent access.</p>



<h2 class="wp-block-heading">What This Means for RPA Teams</h2>



<p class="wp-block-paragraph">If you are running an RPA center of excellence, Pega Agentic Process Fabric introduces a specific strategic question: <strong>where does workflow-governed agent orchestration fit alongside your existing bot estate?</strong></p>



<p class="wp-block-paragraph">The honest answer depends on your starting position. Teams already invested in Pega for case management, customer service, or decisioning have a natural upgrade path — the Agentic Process Fabric extends what they already run. Teams on UiPath, Automation Anywhere, or Power Automate are more likely to encounter Pega as a complementary orchestration layer than a bot replacement, especially if the organization already uses Pega for its core business processes.</p>



<p class="wp-block-paragraph">The more important signal is the pricing model. Flat-fee-per-case pricing directly challenges the token-based economics that have made some enterprises hesitant to scale agentic deployments. If Pega&#8217;s model proves sustainable at scale, expect competitive pressure on other platforms to offer similar predictable-cost options — which would benefit the entire market. For a broader look at how <a href="https://rpabotsworld.com/automation-anywhere-vs-uipath-agentic-procure-to-pay-2026/">Automation Anywhere and UiPath are approaching agentic process automation</a>, see our procure-to-pay comparison.</p>



<h2 class="wp-block-heading">Getting Started</h2>



<p class="wp-block-paragraph">For architects evaluating Pega Agentic Process Fabric, the on-ramp is straightforward:</p>



<ol class="wp-block-list">
<li><strong>Try Blueprint for free</strong> — sign up at <a href="https://www.pega.com/blueprint" target="_blank" rel="noopener nofollow">pega.com/blueprint</a> to explore the AI-assisted workflow design experience</li>



<li><strong>Map your process inventory</strong> — identify which business processes are well-defined enough to model as workflows (Pega&#8217;s strength) versus those that require flexible runtime reasoning</li>



<li><strong>Evaluate MCP compatibility</strong> — if you are building agents on Claude, Gemini, or OpenAI, test interoperability through MCP connections to Pega workflows</li>



<li><strong>Run the AI Token Cost Calculator</strong> — compare Pega&#8217;s flat-fee model against your current or projected token costs on other platforms</li>



<li><strong>Engage an SI partner</strong> — Pega&#8217;s system integrator ecosystem (TCS, Credera, Cognizant, Accenture) can provide proof-of-concept support, especially for legacy modernization scenarios</li>
</ol>



<h2 class="wp-block-heading">Frequently Asked Questions</h2>



<h3 class="wp-block-heading">What is Pega Agentic Process Fabric and how does it differ from regular RPA?</h3>



<p class="wp-block-paragraph">Pega Agentic Process Fabric is an AI agent orchestration layer that coordinates agents, workflows, systems, and data across the enterprise. Unlike traditional RPA, which automates individual screen-level tasks through bots, Agentic Process Fabric orchestrates end-to-end business processes using AI agents governed by predefined workflows. The key difference is scope: RPA bots execute repetitive steps, while the fabric manages entire process outcomes across multiple agents and systems.</p>



<h3 class="wp-block-heading">Does Pega Agentic Process Fabric work with non-Pega AI agents?</h3>



<p class="wp-block-paragraph">Yes. Through Model Context Protocol (MCP) and Agent-to-Agent (A2A) protocol support, the fabric can discover and orchestrate agents from Anthropic Claude, Google Gemini, OpenAI, AWS AgentCore, and other MCP-compatible platforms. Every Pega application becomes an MCP server, enabling third-party agents to invoke Pega workflows without custom integration.</p>



<h3 class="wp-block-heading">How does Pega&#8217;s flat-fee AI pricing work?</h3>



<p class="wp-block-paragraph">Instead of charging per token consumed by LLM calls (which can vary unpredictably based on conversation complexity), Pega charges a flat fee per resolved case. This means the cost of processing a customer service inquiry, a claims adjudication, or a document classification is the same regardless of how many AI interactions the workflow required internally. Pega provides an AI Token Cost Calculator to compare this model against token-based pricing from other platforms.</p>



<h3 class="wp-block-heading">Is Pega Agentic Process Fabric suitable for regulated industries?</h3>



<p class="wp-block-paragraph">It is specifically designed for them. The workflow-first architecture means every agent action maps back to an approved process definition with full audit trails, step-by-step monitoring, and compliance reporting. Agent access is controlled by authentication and role-based permissions, and agent-to-agent interactions are regulated to prevent unauthorized actions.</p>



<h3 class="wp-block-heading">Can Pega Agentic Process Fabric help modernize legacy applications?</h3>



<p class="wp-block-paragraph">Yes, through Pega Blueprint. Organizations can upload legacy application assets — videos of user interactions, screenshots, source code, database schemas — and Blueprint AI will analyze them to produce modernized workflow definitions. This positions the platform as both an agent orchestration layer and a legacy modernization engine.</p>



<h2 class="wp-block-heading">Key Takeaways</h2>



<ul class="wp-block-list">
<li><strong>Workflow-first, not prompt-first:</strong> Pega confines AI reasoning to design time and uses deterministic workflows at runtime — a fundamentally different architectural bet than most agent platforms.</li>



<li><strong>MCP and A2A native:</strong> every Pega application becomes an MCP server, making the platform interoperable with agents from Claude, Gemini, OpenAI, AWS AgentCore, and other MCP-compatible systems.</li>



<li><strong>Flat-fee pricing:</strong> Pega Infinity 26&#8217;s per-resolved-case pricing eliminates token-cost unpredictability — a direct response to the cost overruns that have stalled enterprise agent deployments.</li>



<li><strong>Governance by design:</strong> audit trails, access controls, agent-to-agent regulation, and human-in-the-loop escalation are structural, not optional add-ons.</li>



<li><strong>Legacy modernization built in:</strong> Blueprint AI can reverse-engineer legacy applications into governed agent workflows — a unique capability in the enterprise agent platform market.</li>



<li><strong>Best fit:</strong> regulated industries with complex, well-defined processes and existing Pega investments. Less suited to organizations that need maximum runtime agent flexibility for loosely defined tasks.</li>
</ul>



<h2 class="wp-block-heading">References</h2>



<ol class="wp-block-list">
<li>Pega, &#8220;<a href="https://www.pega.com/about/news/press-releases/pega-agentic-process-fabric-reliably-orchestrates-end-end-ai-automation" target="_blank" rel="noopener nofollow">Pega Agentic Process Fabric Reliably Orchestrates End-to-End AI Automation</a>,&#8221; Press Release, June 2, 2025.</li>



<li>BusinessWire, &#8220;<a href="https://www.businesswire.com/news/home/20260608601073/en/Pega-Powers-AI-Agents-to-Reliably-Drive-Mission-Critical-Work" target="_blank" rel="noopener nofollow">Pega Powers AI Agents to Reliably Drive Mission-Critical Work</a>,&#8221; June 8, 2026.</li>



<li>Constellation Research, &#8220;<a href="https://www.constellationr.com/blog-news/insights/pegasystems-launches-pega-agentic-process-fabric-rides-ai-momentum" target="_blank" rel="noopener nofollow">Pegasystems Launches Pega Agentic Process Fabric, Rides AI Momentum</a>,&#8221; Liz Miller, 2026.</li>



<li>Diginomica, &#8220;<a href="https://diginomica.com/pegas-agentic-approach-puts-workflows-first-prompts-second-heres-why-matters-enterprise-ai-adoption" target="_blank" rel="noopener nofollow">Pega&#8217;s Agentic Approach Puts Workflows First, Prompts Second — Here&#8217;s Why That Matters</a>,&#8221; 2026.</li>



<li>SiliconANGLE, &#8220;<a href="https://siliconangle.com/2025/06/02/pegasystems-expands-features-build-manage-ai-agents/" target="_blank" rel="noopener nofollow">Pegasystems Expands Features to Build and Manage AI Agents</a>,&#8221; June 2, 2025.</li>



<li>DevOps Digest, &#8220;<a href="https://www.devopsdigest.com/pega-infinity-26-released" target="_blank" rel="noopener nofollow">Pega Infinity 26 Released</a>,&#8221; July 2026.</li>



<li>BriefGlance, &#8220;<a href="https://briefglance.com/companies/pegasystems-inc/pulses/58120" target="_blank" rel="noopener nofollow">Pegasystems Launches Pega Infinity 26 with Flat-Fee AI Pricing Model</a>,&#8221; 2026.</li>



<li>Pega, &#8220;<a href="https://www.businesswire.com/news/home/20260714981812/en/" target="_blank" rel="noopener nofollow">Pega Infinity 26 Now Available to Deliver Predictable Outcomes with Predictable AI Costs</a>,&#8221; BusinessWire, July 14, 2026.</li>



<li>CX Today, &#8220;<a href="https://www.cxtoday.com/?p=992594" target="_blank" rel="noopener nofollow">Pegasystems Has Put a Price on Agentic Customer Service. Buyers Need the Fine Print</a>,&#8221; 2026.</li>
</ol>



<p class="wp-block-paragraph"></p>
]]></content:encoded>
					
					<wfw:commentRss>https://rpabotsworld.com/pega-agentic-process-fabric-enterprise-agent-guide/feed/</wfw:commentRss>
			<slash:comments>1</slash:comments>
		
		
		<media:thumbnail url="https://rpabotsworld.com/wp-content/uploads/2026/10/pega-agentic-process-fabric-enterprise-agent-guide-youtube-thumbnail.svg" />	</item>
		<item>
		<title>A2A Joins MCP at the Agentic AI Foundation: The Complete Guide for Agentic AI Architects (2026)</title>
		<link>https://rpabotsworld.com/a2a-joins-mcp-aaif-agentic-ai-architects-guide/</link>
					<comments>https://rpabotsworld.com/a2a-joins-mcp-aaif-agentic-ai-architects-guide/#comments</comments>
		
		<dc:creator><![CDATA[Satish Prasad]]></dc:creator>
		<pubDate>Sat, 05 Sep 2026 17:45:53 +0000</pubDate>
				<category><![CDATA[RPA & Bot Automation]]></category>
		<category><![CDATA[A2A Protocol]]></category>
		<category><![CDATA[MCP]]></category>
		<guid isPermaLink="false">https://rpabotsworld.com/?p=32343</guid>

					<description><![CDATA[A2A joins MCP under the Agentic AI Foundation with 250+ members. Complete architect's guide to the unified protocol stack, v1.0 spec, and production…]]></description>
										<content:encoded><![CDATA[
<p class="wp-block-paragraph">On August 17, 2026, Google&#8217;s Agent2Agent (A2A) protocol formally became a hosted project of the Agentic AI Foundation (AAIF) — the Linux Foundation-directed body that already governs Anthropic&#8217;s Model Context Protocol (MCP), Block&#8217;s goose agent framework, and OpenAI&#8217;s AGENTS.md convention. With more than 250 member organizations including AWS, Microsoft, Anthropic, Google, and OpenAI signed on as governance participants, this move consolidates the two most critical protocol layers of the agentic AI stack under a single neutral roof (<a href="https://aaif.io/blog/a2a-joins-aaif" target="_blank" rel="noopener nofollow">AAIF Official Announcement</a>).</p>



<p class="wp-block-paragraph">If you build multi-agent systems — or you will in the next 12 months — this is not a branding event. It is a structural shift in how your agents will discover, communicate with, and delegate work to agents built by other teams, other vendors, and other frameworks. This guide breaks down exactly what changed, why it matters for enterprise architectures, and what you should do about it now.</p>



<h2 class="wp-block-heading">Table of Contents</h2>



<ul class="wp-block-list">
<li><a href="#what-is-aaif">What Is the Agentic AI Foundation (AAIF)?</a></li>



<li><a href="#what-a2a-does">What A2A Actually Does — And How It Differs from MCP</a></li>



<li><a href="#aaif-stack">The Full AAIF Protocol Stack: Five Layers, One Governance Model</a></li>



<li><a href="#a2a-technical">A2A v1.0 Technical Architecture: Agent Cards, Tasks, and Streaming</a></li>



<li><a href="#production-use">Where A2A Is Already Running in Production</a></li>



<li><a href="#enterprise-impact">What This Means for Enterprise Agentic Architectures</a></li>



<li><a href="#rpa-implications">Implications for RPA Teams Moving to Agentic Workflows</a></li>



<li><a href="#getting-started">Getting Started: How to Adopt A2A + MCP in Your Stack Today</a></li>



<li><a href="#faqs">FAQs</a></li>



<li><a href="#key-takeaways">Key Takeaways</a></li>



<li><a href="#references">References</a></li>
</ul>



<h2 class="wp-block-heading">What Is the Agentic AI Foundation (AAIF)?</h2>



<p class="wp-block-paragraph">The Agentic AI Foundation launched in December 2025 as a Linux Foundation initiative to provide neutral, vendor-independent governance for the infrastructure that powers AI agents in production. Its founding members — Anthropic, OpenAI, and Block — contributed three anchor projects: MCP (the protocol for connecting agents to tools and data), goose (an open-source agent runtime from Block), and AGENTS.md (a specification for communicating instructions and conventions to AI agents).</p>



<p class="wp-block-paragraph">Within eight months, AAIF grew from fewer than 40 members to over 250. Platinum-tier signatories now include AWS, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft, and OpenAI (<a href="https://www.linuxfoundation.org/press/linux-foundation-announces-the-formation-of-the-agentic-ai-foundation" target="_blank" rel="noopener nofollow">Linux Foundation Press Release</a>). That breadth is deliberate. When foundational infrastructure protocols are owned by a single vendor, every downstream team absorbs that vendor&#8217;s roadmap constraints. When they are governed openly, the community shapes what gets built.</p>



<p class="wp-block-paragraph">For context, MCP alone had crossed 97 million monthly SDK downloads (Python + TypeScript combined) by February 2026. The protocol that standardizes how your agents connect to databases, APIs, and enterprise services is no longer any single company&#8217;s project — it belongs to the ecosystem. And now A2A, the protocol that standardizes how your agents talk to <em>each other</em>, sits in the same governance structure.</p>



<h2 class="wp-block-heading">What A2A Actually Does — And How It Differs from MCP</h2>



<p class="wp-block-paragraph">The simplest mental model: <strong>MCP is vertical, A2A is horizontal.</strong></p>



<p class="wp-block-paragraph">MCP standardizes the connection between an agent and its tools. When a Claude-based agent needs to query a PostgreSQL database, read a Google Drive file, or call a Salesforce API, MCP defines the protocol for that interaction — discovery, authentication, invocation, and response. It solves the &#8220;agent-to-tool&#8221; problem. If you&#8217;ve built MCP servers for your internal tools, you&#8217;ve already used this layer. (For a deeper dive, see our guide to <a href="https://rpabotsworld.com/what-is-mcp-server-ai-agents/">what MCP servers are and how they work</a>.)</p>



<p class="wp-block-paragraph">A2A solves a different problem: the &#8220;agent-to-agent&#8221; problem. When a procurement agent built on LangGraph needs to delegate a compliance check to a legal-review agent built on Google ADK, or when a customer-service agent on Salesforce Agentforce needs to route an IT escalation to a ServiceNow agent, A2A defines how those agents discover each other, negotiate capabilities, delegate tasks, and exchange results — regardless of their underlying framework.</p>



<p class="wp-block-paragraph">AAIF CTO Manik Surtani put it directly: &#8220;Where A2A fits is at the collaborative edge; where MCP fits is at the tool integration edge.&#8221; Together they form the plumbing for what analysts increasingly call the agentic AI economy (<a href="https://aaif.io/blog/a2a-joins-aaif" target="_blank" rel="noopener nofollow">AAIF Blog</a>).</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Dimension</th><th>MCP (Model Context Protocol)</th><th>A2A (Agent2Agent)</th></tr></thead><tbody><tr><td><strong>Primary function</strong></td><td>Agent <img src="https://s.w.org/images/core/emoji/17.0.2/72x72/2194.png" alt="↔" class="wp-smiley" style="height: 1em; max-height: 1em;" /> Tool connectivity</td><td>Agent <img src="https://s.w.org/images/core/emoji/17.0.2/72x72/2194.png" alt="↔" class="wp-smiley" style="height: 1em; max-height: 1em;" /> Agent interoperability</td></tr><tr><td><strong>Originated by</strong></td><td>Anthropic (2024)</td><td>Google (April 2025)</td></tr><tr><td><strong>Joined AAIF</strong></td><td>December 2025 (founding project)</td><td>August 2026</td></tr><tr><td><strong>Stable spec</strong></td><td>2026-07-28 stateless spec</td><td>v1.0 (March 2026)</td></tr><tr><td><strong>Core abstraction</strong></td><td>Tools, resources, prompts</td><td>Agent Cards, Tasks, Messages</td></tr><tr><td><strong>Discovery mechanism</strong></td><td>Server manifests</td><td>Agent Cards (signed JSON metadata)</td></tr><tr><td><strong>Communication pattern</strong></td><td>Request/response (tool invocation)</td><td>Task lifecycle (submitted → working → completed) + streaming</td></tr><tr><td><strong>Monthly SDK downloads</strong></td><td>97M+ (Feb 2026)</td><td>Growing rapidly post-v1.0</td></tr><tr><td><strong>Typical use case</strong></td><td>&#8220;My agent needs to read Jira tickets&#8221;</td><td>&#8220;My agent needs to delegate work to your agent&#8221;</td></tr></tbody></table></figure>



<p class="wp-block-paragraph">The critical point: <strong>these protocols are complementary, not competing.</strong> A well-architected multi-agent system uses MCP for each agent&#8217;s tool access and A2A for inter-agent coordination. The AAIF governance model does not merge them — each retains its own technical steering committee — but ensures their roadmaps stay aligned.</p>



<h2 class="wp-block-heading">The Full AAIF Protocol Stack: Five Layers, One Governance Model</h2>



<p class="wp-block-paragraph">With A2A&#8217;s addition, the AAIF now governs a complete five-layer stack for agentic AI infrastructure. Understanding each layer is essential for architects designing production systems.</p>



<h3 class="wp-block-heading">Layer 1: Instructions and Context — AGENTS.md</h3>



<p class="wp-block-paragraph">AGENTS.md standardizes how projects communicate expectations, conventions, and operating instructions to AI agents. Think of it as the &#8220;README for agents&#8221; — a structured document that tells an agent what it can and cannot do within a given project, what coding standards to follow, what tools are available, and what guardrails apply. Contributed by OpenAI.</p>



<h3 class="wp-block-heading">Layer 2: Agent Runtime — goose</h3>



<p class="wp-block-paragraph">Block&#8217;s goose provides the environment in which an agent reasons, plans, invokes capabilities, and executes work. It is an open-source agent framework that handles the execution loop — the &#8220;engine&#8221; that an agent runs on top of.</p>



<h3 class="wp-block-heading">Layer 3: Agent-to-Tool Connectivity — MCP</h3>



<p class="wp-block-paragraph">Anthropic&#8217;s Model Context Protocol standardizes how agents connect to tools, data sources, applications, and services. MCP servers expose capabilities through a uniform interface, and MCP clients (agents) discover and invoke them. The <a href="https://rpabotsworld.com/mcp-2026-07-28-stateless-spec-agentic-ai-guide/">2026-07-28 stateless spec update</a> made MCP servers truly stateless, removing the requirement for persistent connections and making the protocol viable for serverless and edge deployments.</p>



<h3 class="wp-block-heading">Layer 4: Traffic Mediation — agentgateway</h3>



<p class="wp-block-paragraph">agentgateway sits at the boundary between agent systems and infrastructure, handling routing, policy enforcement, and observability. It is the operations layer — where you apply rate limiting, access control, audit logging, and traffic shaping to agent communications.</p>



<h3 class="wp-block-heading">Layer 5: Agent-to-Agent Interoperability — A2A</h3>



<p class="wp-block-paragraph">A2A standardizes how independent agents discover one another, communicate, delegate tasks, and exchange results across systems and organizational boundaries. This is the newest addition and the layer that makes true multi-vendor, multi-framework agent ecosystems possible.</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Layer</th><th>Project</th><th>What It Does</th><th>Contributed By</th></tr></thead><tbody><tr><td>Instructions</td><td>AGENTS.md</td><td>Tells agents how to behave in a project</td><td>OpenAI</td></tr><tr><td>Runtime</td><td>goose</td><td>Executes agent reasoning and planning</td><td>Block</td></tr><tr><td>Tool Access</td><td>MCP</td><td>Connects agents to tools, data, services</td><td>Anthropic</td></tr><tr><td>Operations</td><td>agentgateway</td><td>Routes, monitors, and controls agent traffic</td><td>Community</td></tr><tr><td>Inter-Agent</td><td>A2A</td><td>Agents discover and delegate work to other agents</td><td>Google</td></tr></tbody></table></figure>



<p class="wp-block-paragraph">Every major cloud provider and AI lab is a governance participant. This is not a Google stack or an Anthropic stack — it is the industry stack. For architects, this means you can design multi-agent systems against these protocols with reasonable confidence that they won&#8217;t be deprecated, forked, or locked behind a single vendor&#8217;s paywall.</p>



<h2 class="wp-block-heading">A2A v1.0 Technical Architecture: Agent Cards, Tasks, and Streaming</h2>



<p class="wp-block-paragraph">A2A v1.0 shipped in March 2026 with several critical capabilities that make it production-ready. Here&#8217;s what you need to understand as an implementer.</p>



<h3 class="wp-block-heading">Agent Cards: Discovery and Identity</h3>



<p class="wp-block-paragraph">Every A2A-capable agent publishes an <strong>Agent Card</strong> — a JSON metadata document describing its identity, capabilities, skills, service endpoint, and authentication requirements. Agent Cards are the discovery mechanism: other agents read your card to understand what you can do and how to reach you (<a href="https://a2a-protocol.org/latest/specification/" target="_blank" rel="noopener nofollow">A2A v1.0 Specification</a>).</p>



<p class="wp-block-paragraph">The v1.0 spec added <strong>signed Agent Cards</strong> with cryptographic identity verification. Clients verify the signatures field against the issuer&#8217;s public key (typically held in a curated registry or x5c chain) before trusting the card. This is not optional for enterprise deployments — unsigned agent cards in a multi-vendor system are the equivalent of unsigned API endpoints.</p>



<pre class="wp-block-code"><code>// Simplified Agent Card structure
{
  "@context": "https://a2a-protocol.org/v1.0",
  "name": "compliance-review-agent",
  "description": "Reviews procurement requests against regulatory requirements",
  "url": "https://agents.example.com/compliance",
  "version": "1.0.0",
  "skills": &#91;
    {
      "id": "regulatory-check",
      "name": "Regulatory Compliance Check",
      "description": "Evaluates a procurement request against applicable regulations",
      "inputModes": &#91;"text", "application/json"],
      "outputModes": &#91;"application/json"]
    }
  ],
  "authentication": {
    "schemes": &#91;"oauth2"],
    "credentials": "https://auth.example.com/.well-known/oauth-authorization-server"
  },
  "signatures": { }
}</code></pre>



<h3 class="wp-block-heading">Task Lifecycle: The State Machine</h3>



<p class="wp-block-paragraph">Every interaction in A2A revolves around a strict task state machine. When an agent receives a <code>SendMessage</code> call, it either returns a <code>Message</code> directly (for trivial responses) or instantiates a <code>Task</code> and transitions it through defined states:</p>



<p class="wp-block-paragraph"><strong>submitted → working → input-required → completed | canceled | failed</strong></p>



<p class="wp-block-paragraph">This is fundamentally different from simple request/response patterns. The <code>input-required</code> state is particularly important — it allows agents to negotiate, ask clarifying questions, or request additional context mid-task. A compliance agent reviewing a procurement request can pause in <code>input-required</code>, ask the requesting agent for additional documentation, receive it, and then continue to <code>completed</code>. This models how humans actually collaborate — with back-and-forth — rather than the fire-and-forget pattern of most API integrations.</p>



<h3 class="wp-block-heading">Streaming Architecture</h3>



<p class="wp-block-paragraph">For long-running tasks, A2A supports streaming via Server-Sent Events (SSE). Clients call <code>SendStreamingMessage</code> (to submit and stream in one step) or <code>SubscribeToTask</code> (to attach to an existing task). The server responds with <code>Content-Type: text/event-stream</code> and pushes <code>TaskStatusUpdateEvent</code> and <code>TaskArtifactUpdateEvent</code> objects until the task reaches a terminal state.</p>



<p class="wp-block-paragraph">This means your orchestrating agent doesn&#8217;t need to poll. It opens a stream, processes status updates and partial results as they arrive, and can display progress to end users in real time.</p>



<h3 class="wp-block-heading">Multi-Tenancy and Version Negotiation</h3>



<p class="wp-block-paragraph">The v1.0 spec added two enterprise-critical capabilities. <strong>Multi-tenancy</strong> allows a single A2A server to serve multiple tenants with isolated task namespaces. <strong>Version negotiation</strong> allows agents running different spec versions to discover the highest mutually supported version and communicate at that level — essential for organizations rolling out upgrades across a fleet of agents at different cadences.</p>



<h2 class="wp-block-heading">Where A2A Is Already Running in Production</h2>



<p class="wp-block-paragraph">A2A is not a whitepaper protocol. As of mid-2026, it is running in production across mobile platforms, cloud infrastructure, financial services, and supply chain logistics. Here are the most significant deployments (<a href="https://www.linuxfoundation.org/press/a2a-protocol-surpasses-150-organizations-lands-in-major-cloud-platforms-and-sees-enterprise-production-use-in-first-year" target="_blank" rel="noopener nofollow">Linux Foundation Press Release, April 2026</a>).</p>



<h3 class="wp-block-heading">Huawei HarmonyOS: Mobile-Scale Agent Orchestration</h3>



<p class="wp-block-paragraph">Huawei standardized A2A as the protocol between Celia (its OS-level AI assistant) and in-app agents across the HarmonyOS developer platform. Supported scenarios include Celia delegating long-running tasks to app agents, Celia controlling app UI through an agent, and Celia requesting contextual recommendations from application agents. This is agent-to-agent communication operating at <em>mobile operating system scale</em> — billions of potential interactions daily.</p>



<p class="wp-block-paragraph">Tencent&#8217;s WeChat is among the first major apps integrating with Huawei and other Android OEM assistants over A2A, enabling messages, voice calls, and video calls initiated through an AI assistant. The flow runs with dual authorization through the A2A protocol.</p>



<h3 class="wp-block-heading">PayPal: Agentic Commerce</h3>



<p class="wp-block-paragraph">PayPal deployed A2A in production for merchant-facing workflows. A sales agent receives a natural-language request, uses A2A to locate and authenticate a PayPal-provided payment agent via its Agent Card, and initiates a transaction. Google Cloud and PayPal are extending this into the Agent Payments Protocol (AP2), where shopping and merchant agents communicate over A2A throughout product discovery, pricing, and order fulfillment, with AP2 handling the payment authorization layer. Over 60 organizations across payments and financial services already support this initiative.</p>



<h3 class="wp-block-heading">Cloud Platform Native Integration</h3>



<p class="wp-block-paragraph">Every major cloud platform now supports A2A natively:</p>



<ul class="wp-block-list">
<li><strong>Google Cloud:</strong> Supports developing and deploying A2A agents through Agent Development Kit (ADK), Agent Engine, Cloud Run, and GKE. If you&#8217;ve been <a href="https://rpabotsworld.com/building-multi-agent-systems-with-google-adk-the-complete-step-by-step-guide/">building multi-agent systems with Google ADK</a>, A2A support is already built into the framework.</li>



<li><strong>Microsoft Azure:</strong> Azure AI Foundry lets agents expose A2A endpoints and discover external agents through standard discovery. This integrates directly with Copilot Studio&#8217;s multi-agent orchestration capabilities.</li>



<li><strong>AWS:</strong> Amazon Bedrock AgentCore can host and operate A2A servers, enabling cross-framework and cross-cloud agent communication.</li>
</ul>



<h3 class="wp-block-heading">Enterprise IT: Cross-Vendor Agent Collaboration</h3>



<p class="wp-block-paragraph">A Salesforce CRM agent can route a support escalation to a ServiceNow ITSM agent. A Google Workspace agent can delegate document summarization to a specialized LLM agent on Azure. These aren&#8217;t theoretical scenarios — they are the explicit design targets of the protocol and the reason companies like Salesforce, SAP, ServiceNow, and Workday are all supporters.</p>



<h2 class="wp-block-heading">What This Means for Enterprise Agentic Architectures</h2>



<p class="wp-block-paragraph">The consolidation of A2A and MCP under one governance body has several concrete implications for architects designing production agent systems.</p>



<h3 class="wp-block-heading">1. The End of Custom Integration Code Between Agents</h3>



<p class="wp-block-paragraph">Before A2A, connecting agents built on different frameworks required custom integration code for every pairing. A LangGraph agent talking to a Google ADK agent needed bespoke glue. A UiPath process invoking a Bedrock agent needed a custom connector. With A2A, any agent that publishes an Agent Card and implements the task lifecycle can communicate with any other A2A-compatible agent. The cost isn&#8217;t in the agents — it&#8217;s in eliminating the O(n²) integration matrix.</p>



<h3 class="wp-block-heading">2. Protocol Vendor Lock-in Risk Drops Dramatically</h3>



<p class="wp-block-paragraph">When MCP was Anthropic&#8217;s project and A2A was Google&#8217;s project, architects legitimately worried about betting on one vendor&#8217;s protocol. With both under AAIF governance — and with every major cloud provider and AI lab signed on — the risk profile changes fundamentally. These are now industry standards with the same governance model that governs Linux, Kubernetes, and other foundational infrastructure. As our <a href="https://rpabotsworld.com/servicenow-ai-control-tower-enterprise-agent-governance-guide/">guide to enterprise agent governance</a> explores, governance is the hardest problem in agent deployments — and having protocol governance settled at the industry level removes one category of that risk entirely.</p>



<h3 class="wp-block-heading">3. The Agent Card Becomes Your Agent&#8217;s Resume</h3>



<p class="wp-block-paragraph">Agent Cards are not just a technical artifact. They are the mechanism by which your agents advertise their capabilities in a marketplace of agents. Organizations that invest in clear, well-structured Agent Cards — with precise skill descriptions, proper authentication, and cryptographic signatures — will find their agents get discovered and delegated to. Those that don&#8217;t will build agents that sit in isolation.</p>



<h3 class="wp-block-heading">4. Multi-Cloud Agent Architectures Become First-Class</h3>



<p class="wp-block-paragraph">With AWS, Azure, and Google Cloud all supporting A2A natively, you can now design architectures where agents on different clouds collaborate seamlessly. A fraud-detection agent on AWS Bedrock can delegate to a customer-history agent on Azure AI Foundry, which can invoke a risk-scoring agent on Google Cloud&#8217;s Agent Engine — all through standardized A2A communication. This is a concrete answer to the multi-cloud agent challenge that enterprises have been asking about.</p>



<h3 class="wp-block-heading">5. IBM&#8217;s ACP Merger Validates the Convergence</h3>



<p class="wp-block-paragraph">In August 2025, IBM&#8217;s Agent Communication Protocol (ACP) merged into A2A. This was an early signal that the industry was consolidating around A2A rather than sustaining competing protocols. IBM — the company behind <a href="https://rpabotsworld.com/ibm-watsonx-orchestrate-agentic-control-plane-guide/">watsonx Orchestrate&#8217;s agentic control plane</a> — decided to contribute its protocol work into A2A rather than maintain a separate standard. When a company with IBM&#8217;s enterprise footprint makes that call, it tells you where the industry consensus landed.</p>



<h2 class="wp-block-heading">Implications for RPA Teams Moving to Agentic Workflows</h2>



<p class="wp-block-paragraph">If your team currently runs UiPath, Automation Anywhere, Blue Prism, or Power Automate automations, the A2A + MCP stack has direct implications for your transition path to agentic workflows.</p>



<h3 class="wp-block-heading">RPA Processes as A2A-Accessible Agents</h3>



<p class="wp-block-paragraph">The A2A protocol makes it possible to wrap existing RPA processes as agent-accessible services. An RPA bot that processes invoices in SAP can be exposed via an Agent Card, making it discoverable and callable by any A2A-compatible agent. The bot doesn&#8217;t need to be rebuilt — it needs a wrapper that publishes its capabilities and implements the task lifecycle.</p>



<p class="wp-block-paragraph">This is significant because it provides a migration path that doesn&#8217;t require rip-and-replace. Your existing RPA investments become nodes in a broader agentic network. The <a href="https://rpabotsworld.com/uipath-vs-automation-anywhere-vs-blue-prism-agentic-platforms-2026/">latest comparison of UiPath, Automation Anywhere, and Blue Prism&#8217;s agentic capabilities</a> shows that all three vendors are moving toward agent-oriented architectures — A2A gives them a standard protocol to target.</p>



<h3 class="wp-block-heading">MCP for Tool Access, A2A for Process Orchestration</h3>



<p class="wp-block-paragraph">Here&#8217;s a practical architecture pattern for RPA-to-agentic migration:</p>



<ul class="wp-block-list">
<li><strong>MCP servers</strong> wrap your existing integrations: SAP connections, database queries, file system operations, email interactions. Each becomes a standard MCP tool that any agent can discover and invoke.</li>



<li><strong>A2A</strong> handles the orchestration layer: a customer-onboarding agent delegates KYC checks to a compliance agent, which delegates document extraction to an IDP agent, which delegates data entry to an RPA-wrapped SAP agent.</li>
</ul>



<p class="wp-block-paragraph">This pattern preserves your existing automation logic (the RPA bots doing the actual work) while adding an intelligent orchestration layer on top. It is a practical answer to the question every RPA team is asking: &#8220;How do we get to agentic without throwing away what we&#8217;ve built?&#8221;</p>



<h3 class="wp-block-heading">The Governance Question</h3>



<p class="wp-block-paragraph">Enterprise RPA programs already have governance frameworks — change management, access control, audit trails, error handling. The A2A protocol maps cleanly to these requirements. Agent Cards include authentication requirements. The task lifecycle provides auditable state transitions. The <a href="https://rpabotsworld.com/openai-presence-enterprise-ai-agent-platform-guide/">enterprise agent deployment considerations</a> we&#8217;ve covered previously apply directly — A2A doesn&#8217;t replace your governance; it gives it a standard protocol to enforce against.</p>



<h2 class="wp-block-heading">Getting Started: How to Adopt A2A + MCP in Your Stack Today</h2>



<p class="wp-block-paragraph">Here&#8217;s a practical roadmap for teams that want to start building with both protocols.</p>



<h3 class="wp-block-heading">Step 1: Start with MCP (If You Haven&#8217;t Already)</h3>



<p class="wp-block-paragraph">MCP has a larger ecosystem and more immediate utility for most teams. Wrap your most-used internal tools as MCP servers. The Python and TypeScript SDKs are mature, and most agent frameworks (LangGraph, Google ADK, Claude, Bedrock Agents) support MCP natively. Start with your database, your CRM, and your document store.</p>



<h3 class="wp-block-heading">Step 2: Design Your Agent Card</h3>



<p class="wp-block-paragraph">Before you implement A2A, design the Agent Card for your most valuable agent. What skills does it expose? What input/output modes does it support? What authentication is required? This exercise forces you to think about your agent&#8217;s external interface — which is often underdefined in internal agent projects.</p>



<h3 class="wp-block-heading">Step 3: Choose Your A2A SDK</h3>



<p class="wp-block-paragraph">Google provides official SDKs for Python, TypeScript/JavaScript, Java, and Go. The SDKs handle the task lifecycle, streaming, and Agent Card serving. If you&#8217;re on Google ADK, A2A support is built in. If you&#8217;re on LangGraph or another framework, use the standalone SDK.</p>



<pre class="wp-block-code"><code># Python: Minimal A2A server using the official SDK
from a2a.server import A2AServer, AgentCard, Skill

card = AgentCard(
    name="invoice-processor",
    description="Processes and validates supplier invoices",
    url="https://agents.internal.example.com/invoice",
    skills=&#91;
        Skill(
            id="validate-invoice",
            name="Invoice Validation",
            description="Validates invoice against PO and contract terms",
            input_modes=&#91;"application/json"],
            output_modes=&#91;"application/json"]
        )
    ]
)

server = A2AServer(card=card)

@server.on_message
async def handle(message, context):
    # Your agent logic here
    invoice_data = message.content
    result = await validate_against_po(invoice_data)
    return context.complete(result)

server.run(port=8080)</code></pre>



<h3 class="wp-block-heading">Step 4: Test Cross-Framework Communication</h3>



<p class="wp-block-paragraph">The value of A2A appears when agents on different frameworks need to collaborate. Build a simple two-agent system where Agent A (on one framework) delegates a task to Agent B (on another) via A2A. Verify the full lifecycle: discovery via Agent Card, task submission, status updates, and result retrieval. This is your proof of concept for multi-vendor agent architectures.</p>



<h3 class="wp-block-heading">Step 5: Layer in agentgateway for Operations</h3>



<p class="wp-block-paragraph">Once you have agents communicating via A2A, add AAIF&#8217;s agentgateway for routing, policy enforcement, and observability. This is where you apply rate limiting, access control, and audit logging — the operations layer that makes agent communication production-ready.</p>



<h2 class="wp-block-heading">Frequently Asked Questions</h2>



<h3 class="wp-block-heading">Does A2A replace MCP?</h3>



<p class="wp-block-paragraph">No. They solve different problems and are designed to work together. MCP handles agent-to-tool connectivity (your agent querying a database or calling an API). A2A handles agent-to-agent communication (your agent delegating work to another agent). A production system typically uses both — MCP for each agent&#8217;s tool access and A2A for inter-agent orchestration.</p>



<h3 class="wp-block-heading">Do I need to switch agent frameworks to use A2A?</h3>



<p class="wp-block-paragraph">No. A2A is framework-agnostic by design. An agent built on LangGraph can communicate with an agent built on Google ADK, Claude, or a custom framework. You add A2A support to your existing agent — you don&#8217;t rebuild the agent. The official SDKs (Python, TypeScript, Java, Go) handle the protocol layer.</p>



<h3 class="wp-block-heading">Is A2A ready for production use?</h3>



<p class="wp-block-paragraph">Yes. A2A v1.0 shipped in March 2026 with signed Agent Cards, multi-tenancy, and version negotiation. It is running in production at Huawei (mobile OS scale), PayPal (agentic commerce), and across AWS, Azure, and Google Cloud. Over 150 organizations are active supporters.</p>



<h3 class="wp-block-heading">How does A2A handle security and authentication?</h3>



<p class="wp-block-paragraph">Agent Cards include authentication requirements (OAuth 2.0, API keys, or custom schemes). The v1.0 spec added cryptographic signatures for Agent Cards, so clients can verify an agent&#8217;s identity before trusting its card. The task lifecycle is fully auditable, with each state transition logged.</p>



<h3 class="wp-block-heading">Can existing RPA processes be exposed via A2A?</h3>



<p class="wp-block-paragraph">Yes. An RPA process can be wrapped with an A2A server that publishes an Agent Card describing the process&#8217;s capabilities and implements the task lifecycle. The RPA bot continues to execute the actual work — A2A provides the discovery and communication layer that makes it accessible to other agents.</p>



<h2 class="wp-block-heading">Key Takeaways</h2>



<ul class="wp-block-list">
<li><strong>A2A + MCP under one roof is a structural shift.</strong> The two most critical agent protocol layers — tool connectivity (MCP) and inter-agent communication (A2A) — are now governed by the same Linux Foundation body with 250+ members including every major cloud provider and AI lab.</li>



<li><strong>MCP is vertical, A2A is horizontal.</strong> Use MCP to connect agents to tools and data. Use A2A to connect agents to other agents. Both are complementary and both are needed for production multi-agent systems.</li>



<li><strong>A2A is already in production.</strong> Huawei (mobile OS), PayPal (commerce), and all three major cloud platforms run A2A today. This is not vaporware.</li>



<li><strong>RPA teams have a migration path.</strong> Existing RPA processes can be wrapped as A2A-accessible agents without rebuilding them, preserving automation investments while adding intelligent orchestration.</li>



<li><strong>Agent Cards are your agent&#8217;s external interface.</strong> Invest in clear, well-structured Agent Cards with proper authentication and signatures. They determine whether your agents get discovered and trusted in multi-vendor ecosystems.</li>



<li><strong>Start with MCP, add A2A when you need multi-agent.</strong> For most teams, wrapping internal tools as MCP servers delivers immediate value. Layer A2A when you need agents to delegate work across framework or vendor boundaries.</li>
</ul>



<h2 class="wp-block-heading">References</h2>



<ol class="wp-block-list">
<li>Agentic AI Foundation. &#8220;A2A joins AAIF&#8217;s open agentic stack.&#8221; August 17, 2026. <a href="https://aaif.io/blog/a2a-joins-aaif" target="_blank" rel="noopener nofollow">https://aaif.io/blog/a2a-joins-aaif</a></li>



<li>Forbes. &#8220;Agent2Agent Joins The Agentic AI Foundation Alongside MCP.&#8221; August 19, 2026. <a href="https://www.forbes.com/sites/janakirammsv/2026/08/19/agent2agent-joins-the-agentic-ai-foundation-alongside-mcp/" target="_blank" rel="noopener nofollow">https://www.forbes.com/sites/janakirammsv/2026/08/19/</a></li>



<li>Axios. &#8220;AI agents inch toward interoperability.&#8221; August 17, 2026. <a href="https://www.axios.com/2026/08/17/a2a-agentic-ai-foundation-open-ai-standards" target="_blank" rel="noopener nofollow">https://www.axios.com/2026/08/17/</a></li>



<li>Linux Foundation. &#8220;A2A Protocol Surpasses 150 Organizations.&#8221; April 2026. <a href="https://www.linuxfoundation.org/press/a2a-protocol-surpasses-150-organizations-lands-in-major-cloud-platforms-and-sees-enterprise-production-use-in-first-year" target="_blank" rel="noopener nofollow">https://www.linuxfoundation.org/press/</a></li>



<li>Linux Foundation. &#8220;Announces the Formation of the Agentic AI Foundation (AAIF).&#8221; December 2025. <a href="https://www.linuxfoundation.org/press/linux-foundation-announces-the-formation-of-the-agentic-ai-foundation" target="_blank" rel="noopener nofollow">https://www.linuxfoundation.org/press/</a></li>



<li>A2A Protocol. &#8220;v1.0 Specification.&#8221; March 2026. <a href="https://a2a-protocol.org/latest/specification/" target="_blank" rel="noopener nofollow">https://a2a-protocol.org/latest/specification/</a></li>



<li>AI Magazine. &#8220;Why Did Google&#8217;s A2A Join the Agentic AI Foundation?&#8221; August 2026. <a href="https://aimagazine.com/news/why-did-googles-a2a-join-the-agentic-ai-foundation" target="_blank" rel="noopener nofollow">https://aimagazine.com/news/</a></li>



<li>Tyk. &#8220;A2A protocol: Architecture and technical specification.&#8221; 2026. <a href="https://tyk.io/learning-center/a2a-protocol-architecture-and-technical-specification/" target="_blank" rel="noopener nofollow">https://tyk.io/learning-center/</a></li>



<li>Pebblous. &#8220;A2A Joins MCP at the Agentic AI Foundation.&#8221; 2026. <a href="https://blog.pebblous.ai/blog/a2a-mcp-agentic-ai-foundation-authorization/en/" target="_blank" rel="noopener nofollow">https://blog.pebblous.ai/blog/</a></li>



<li>Zylos Research. &#8220;Agent Interoperability Protocols 2026: MCP, A2A, ACP and the Path to Convergence.&#8221; March 2026. <a href="https://zylos.ai/research/2026-03-26-agent-interoperability-protocols-mcp-a2a-acp-convergence/" target="_blank" rel="noopener nofollow">https://zylos.ai/research/</a></li>
</ol>
]]></content:encoded>
					
					<wfw:commentRss>https://rpabotsworld.com/a2a-joins-mcp-aaif-agentic-ai-architects-guide/feed/</wfw:commentRss>
			<slash:comments>24</slash:comments>
		
		
			</item>
		<item>
		<title>UiPath Autopilot Is Now a Coding Agent: What the August 2026 GA Means for RPA Teams</title>
		<link>https://rpabotsworld.com/uipath-autopilot-coding-agent-ga-guide-2/</link>
					<comments>https://rpabotsworld.com/uipath-autopilot-coding-agent-ga-guide-2/#comments</comments>
		
		<dc:creator><![CDATA[Satish Prasad]]></dc:creator>
		<pubDate>Wed, 19 Aug 2026 02:51:09 +0000</pubDate>
				<category><![CDATA[RPA & Bot Automation]]></category>
		<category><![CDATA[UiPath]]></category>
		<category><![CDATA[MCP]]></category>
		<guid isPermaLink="false">https://rpabotsworld.com/?p=32314</guid>

					<description><![CDATA[UiPath Autopilot is now a GA coding agent in Studio. How the skills architecture, MCP support, and UiPath for Coding Agents platform reshape enterprise RPA…]]></description>
										<content:encoded><![CDATA[
<p class="wp-block-paragraph">On August 12, 2026, UiPath shipped what may be the most consequential update in its platform&#8217;s history: <a href="https://docs.uipath.com/autopilot/other/latest/release-notes/august-2026" rel="nofollow noopener" target="_blank">Autopilot became a generally available coding agent</a> inside Studio Desktop. Not a copilot that suggests snippets. Not a chatbot bolted onto the IDE. A full coding agent that plans automations from specification documents, builds workflows that follow UiPath conventions, runs them, diagnoses failures, and restructures code with better error handling — all through natural language conversation.</p>



<p class="wp-block-paragraph">This isn&#8217;t a feature release. It&#8217;s a platform identity shift. UiPath is no longer designing primarily for human developers dragging activities onto a canvas. The primary consumer of its platform infrastructure is now a coding agent, and the human&#8217;s role becomes describing intent, exercising judgment, and approving what the agent produces.</p>



<p class="wp-block-paragraph">For RPA teams that have spent years building competency around Studio&#8217;s visual designer, this raises urgent questions: What exactly changed? How does the skills-based architecture work? What does &#8220;UiPath for Coding Agents&#8221; mean for teams already running Claude Code or Codex? And what does the governance model actually cover?</p>



<p class="wp-block-paragraph">This guide breaks down everything practitioners need to know — the architecture, the capabilities, the licensing, the gaps, and the strategic implications for enterprise automation programs.</p>



<h2 class="wp-block-heading">Table of Contents</h2>



<ul class="wp-block-list">
<li><a href="#what-shipped">What Shipped on August 12</a></li>



<li><a href="#skills-architecture">The Skills-Based Architecture: How It Actually Works</a></li>



<li><a href="#coding-agents-platform">UiPath for Coding Agents: The Broader Platform Play</a></li>



<li><a href="#operate-ga">Operate with Coding Agents: The Other August GA</a></li>



<li><a href="#what-autopilot-can-do">What Autopilot Can Actually Do Now</a></li>



<li><a href="#governance-model">The Governance Model — and Its Gaps</a></li>



<li><a href="#licensing">Licensing and Availability</a></li>



<li><a href="#what-changes-for-rpa-teams">What This Changes for RPA Teams</a></li>



<li><a href="#comparison-table">Autopilot Coding Agent vs. Traditional Studio Development</a></li>



<li><a href="#strategic-implications">Strategic Implications: Dines&#8217; Platform Bet</a></li>



<li><a href="#faqs">FAQs</a></li>



<li><a href="#key-takeaways">Key Takeaways</a></li>



<li><a href="#references">References</a></li>
</ul>



<h2 class="wp-block-heading">What Shipped on August 12</h2>



<p class="wp-block-paragraph">Autopilot&#8217;s coding agent capability reached general availability in Studio Desktop STS, starting with build 2026.0.199. The update transforms Autopilot from the earlier assistant model — which offered search, expression fixes, and guided suggestions — into a full agentic loop that can plan, execute, observe results, and self-correct across the entire automation lifecycle.</p>



<p class="wp-block-paragraph">Three things make this release architecturally distinct from what came before:</p>



<p class="wp-block-paragraph"><strong>First, the skills-and-tools runtime.</strong> Rather than running on a fixed pipeline where each capability is hardcoded, Autopilot operates on a modular system of skills, tools, and commands. UiPath ships over 30 built-in tools and an auto-loaded skills catalog that adapts to your current task context. All tools are toggleable from the connected sources panel — you can enable or disable capabilities depending on what you&#8217;re building. This is the same skills infrastructure that powers the broader &#8220;UiPath for Coding Agents&#8221; platform, which means Autopilot and third-party coding agents (Claude Code, Codex) share the same underlying capability layer.</p>



<p class="wp-block-paragraph"><strong>Second, MCP server support.</strong> Autopilot can connect to <a href="https://rpabotsworld.com/what-is-mcp-server-ai-agents/">Model Context Protocol servers</a> configured in Orchestrator, opening external service integrations without custom code. This is significant because it means the agent isn&#8217;t limited to UiPath&#8217;s own integration surface — any tool exposed through MCP becomes available during the build-and-operate cycle.</p>



<p class="wp-block-paragraph"><strong>Third, AGENTS.md support.</strong> Teams can capture project conventions — coding standards, naming patterns, error-handling requirements, selector strategies — in an AGENTS.md file at the project root. Autopilot reads this file at session start and follows those conventions for the duration of its work. This is the same open format used by Claude Code and other agents, which means project-level governance travels with the repo regardless of which agent is driving.</p>



<h2 class="wp-block-heading">The Skills-Based Architecture: How It Actually Works</h2>



<p class="wp-block-paragraph">The shift from a fixed pipeline to a skills-based runtime is the most important architectural change in this release, and it&#8217;s worth understanding in detail because it determines what the agent can and cannot do.</p>



<p class="wp-block-paragraph">In UiPath&#8217;s model, a <strong>skill</strong> is a task-oriented instruction bundle — not an MCP server, not a plugin, not a traditional activity package. <a href="https://docs.uipath.com/uipath-cli/standalone/latest/user-guide/concepts-skills" rel="nofollow noopener" target="_blank">Skills are published by UiPath</a> and describe, in natural language enriched with structured metadata, how to accomplish a specific class of tasks using the <code>uip</code> CLI. When a coding agent has UiPath skills installed, it knows when to pack a Solution, how to chain <code>publish</code> with <code>deploy</code> and <code>run</code>, when to wait for a job, and how to inspect an Orchestrator folder.</p>



<p class="wp-block-paragraph">The key design decision: skills are not MCP servers. UiPath&#8217;s documentation is explicit about this. The skills system is the primary AI integration path in the current 1.x release. MCP is available for specialized setups — and Autopilot supports it — but skills are how UiPath teaches agents to use its platform.</p>



<p class="wp-block-paragraph">This matters for two reasons. First, skills can encode multi-step workflows with conditional logic (&#8220;if this deploy fails, check the package version and retry with the correct dependency&#8221;), not just expose atomic tool calls. Second, because skills are instruction bundles rather than API surfaces, they can be authored by anyone — UiPath publishes official ones, but you can write your own and reference UiPath&#8217;s skills inside them. A team could, for example, write a skill that enforces their specific deployment approval process on top of UiPath&#8217;s standard publish-deploy chain.</p>



<p class="wp-block-paragraph">The auto-loaded skills catalog is context-aware: it surfaces different skills depending on whether you&#8217;re building an RPA workflow, operating Orchestrator, or troubleshooting a failed job. Over 30 built-in tools handle the mechanical work — file operations, project scaffolding, selector manipulation, expression evaluation — while skills handle the higher-level reasoning about which tools to chain and in what order.</p>



<h2 class="wp-block-heading">UiPath for Coding Agents: The Broader Platform Play</h2>



<p class="wp-block-paragraph">Autopilot&#8217;s GA is the native implementation, but the bigger strategic move launched three months earlier. On May 12, 2026, UiPath announced <a href="https://www.uipath.com/newsroom/uipath-for-coding-agents-launch" rel="nofollow noopener" target="_blank">UiPath for Coding Agents</a> — platform-wide integration enabling any coding agent to become enterprise-deployable. Initial support covers Claude Code (Anthropic) and OpenAI Codex, with additional integrations planned through 2026.</p>



<p class="wp-block-paragraph">The architecture is deliberately vendor-neutral. As <a href="https://diginomica.com/uipath-opens-its-platform-every-coding-agent-heres-why-claude-code-and-codex-go-first" rel="nofollow noopener" target="_blank">diginomica&#8217;s analysis</a> noted, UiPath made a strategic decision not to build its own coding agent — instead, it built the platform layer that every coding agent needs to operate in an enterprise. Claude Code can run in one department, Codex in another, and a future agent slots in alongside without re-platforming. The orchestration layer is the constant.</p>



<p class="wp-block-paragraph">This addresses a real enterprise pain point. Coding agents in 2026 are powerful but isolated. They produce impressive demos inside development sandboxes, but connecting their output to CI/CD pipelines, code review processes, security policies, credential vaults, and production deployment workflows requires manual handoffs at almost every step. UiPath&#8217;s pitch is that Maestro — its workflow orchestrator built on Temporal&#8217;s durable execution technology — provides the observability, execution, and governance scaffolding regardless of which agent generated the underlying automation.</p>



<p class="wp-block-paragraph">For existing UiPath customers, the math is straightforward: their Orchestrator, credential stores, RBAC policies, audit trails, and runtime infrastructure work with agent-generated automations exactly the same way they work with human-generated ones. For teams evaluating coding agents for the first time, UiPath is positioning itself as the answer to &#8220;now that the agent wrote the code, how do I actually deploy, govern, and operate it at scale?&#8221;</p>



<p class="wp-block-paragraph">Daniel Dines, UiPath&#8217;s CEO, framed the shift directly in the <a href="https://www.uipath.com/newsroom/uipath-for-coding-agents-launch" rel="nofollow noopener" target="_blank">press release</a>: &#8220;The emergence of coding agents signals a fundamental shift in the definition of a builder on our platform. We are first to market with a platform that treats AI-generated automations as first-class citizens, with the same governance, reliability, and scale that enterprises demand.&#8221;</p>



<h2 class="wp-block-heading">Operate with Coding Agents: The Other August GA</h2>



<p class="wp-block-paragraph">Nine days before Autopilot&#8217;s coding agent GA, a less-noticed but equally important milestone hit: <a href="https://docs.uipath.com/coding-agents/standalone/latest/release-notes/august-2026" rel="nofollow noopener" target="_blank">Operating UiPath with a coding agent reached general availability on August 3, 2026</a>.</p>



<p class="wp-block-paragraph">This is the operational counterpart to the build capability. A coding agent with UiPath skills installed can now drive most operational tasks across the UiPath platform — managing folders, triggering jobs, working with queues, configuring assets, pulling audit logs, managing connections — through natural-language conversation.</p>



<p class="wp-block-paragraph">The capability spans what UiPath calls the &#8220;operational surface&#8221; of the platform: access and identity, runtime infrastructure, deployment, execution, data and configuration, and oversight. You state the outcome rather than the command, and the agent picks the right <code>uip</code> command from the installed skills. Crucially, the agent reads on its own judgment but confirms with you before anything that creates, updates, or deletes.</p>



<p class="wp-block-paragraph">For RPA teams that currently operate through Orchestrator&#8217;s web UI or PowerShell scripts, this collapses the operational workflow into conversational interaction. Instead of clicking through Orchestrator to find a failed job, checking its logs, identifying the root cause, fixing the automation, and redeploying, you describe the problem and the agent walks the entire chain — potentially including the fix and redeployment — with human approval at each destructive step.</p>



<h2 class="wp-block-heading">What Autopilot Can Actually Do Now</h2>



<p class="wp-block-paragraph">The August 2026 release documentation breaks Autopilot&#8217;s capabilities into six functional areas, each worth examining for what it means in practice:</p>



<h3 class="wp-block-heading">Plan from a Spec</h3>



<p class="wp-block-paragraph">Hand Autopilot a Process Definition Document (PDD) or Solution Design Document (SDD) and it generates an automation plan, then builds the implementation. This isn&#8217;t template matching — the agent reads the document, identifies the process steps, maps them to UiPath activities and patterns, and produces a working project structure. For teams with existing PDD libraries, this turns documentation backlog into automation backlog overnight.</p>



<h3 class="wp-block-heading">Build and Edit</h3>



<p class="wp-block-paragraph">Autopilot generates and modifies both XAML workflows and coded (.cs) automations that follow UiPath conventions. It handles UI automation directly from prompts — describe what you need extracted or interacted with, and the agent builds the selectors, error handling, and retry logic. It can also edit existing workflows: point it at a legacy automation and ask for improvements, and it restructures the code with updated patterns.</p>



<h3 class="wp-block-heading">Extract from UIs</h3>



<p class="wp-block-paragraph">Data extraction through UI automation, end to end from a prompt. Describe the application and the data you need, and Autopilot builds the selectors, handles the navigation, and structures the output. This is particularly useful for the long tail of enterprise applications that lack APIs — the systems that RPA was originally built to handle.</p>



<h3 class="wp-block-heading">Run and Troubleshoot</h3>



<p class="wp-block-paragraph">Autopilot runs the automation it builds, observes the results, and troubleshoots failures — including deployed jobs that exhibit flaky behavior in production. This closes the build-test-debug loop inside a single agent session rather than requiring the developer to switch between Studio, Orchestrator, and log files.</p>



<h3 class="wp-block-heading">Debug and Fix</h3>



<p class="wp-block-paragraph">Root cause analysis and automated fix proposals, including broken selectors. Selector debugging has always been one of the most time-consuming aspects of RPA development — applications update their UI, selectors break, and developers spend hours in UiExplorer trying to build resilient alternatives. Autopilot handles this programmatically, proposing fixes that use hardened Object Repository selectors.</p>



<h3 class="wp-block-heading">Explain and Improve</h3>



<p class="wp-block-paragraph">Walk-throughs of unfamiliar automations, automatic documentation generation, and code restructuring with better error handling, logging, and selector strategies. For teams inheriting automation portfolios — through acquisitions, team changes, or vendor transitions — this converts opaque legacy workflows into documented, maintainable code.</p>



<h2 class="wp-block-heading">The Governance Model — and Its Gaps</h2>



<p class="wp-block-paragraph">UiPath&#8217;s governance story for coding agents has two parts, and the boundary between them matters for regulated enterprises.</p>



<p class="wp-block-paragraph"><strong>What&#8217;s governed today:</strong> Every automation entering the platform — whether built by a human or a coding agent — goes through the same governance layer. Policy enforcement, audit trails, credential vaults, RBAC, and runtime controls are standard. This is UiPath&#8217;s existing enterprise infrastructure, and it works the same way regardless of how the automation was created. The orchestration layer (Maestro) persists the state of every workflow step using Temporal&#8217;s durable execution, so automations survive infrastructure failures and can be paused, resumed, and audited end to end.</p>



<p class="wp-block-paragraph"><strong>What&#8217;s not yet governed:</strong> As <a href="https://diginomica.com/uipath-opens-its-platform-every-coding-agent-heres-why-claude-code-and-codex-go-first" rel="nofollow noopener" target="_blank">diginomica&#8217;s Alyx MacQueen noted</a> in her analysis, the governance model covers the output — the automation artifact that enters the platform — but the question of what the coding agent does <em>before</em> submission remains open. The prompting, the reasoning trace, the iterations during code generation, the credentials the agent touches during the build process — these are not yet part of the governance surface.</p>



<p class="wp-block-paragraph">For teams in financial services, healthcare, or government, this distinction matters. Regulatory frameworks increasingly require auditability not just of what was deployed, but of the process that produced it. If a coding agent generated an automation that processes patient data, auditors may want to see the reasoning chain that led to the design decisions — not just the final XAML file that passed through Orchestrator&#8217;s approval workflow.</p>



<p class="wp-block-paragraph">UiPath has acknowledged this gap implicitly through the AGENTS.md support: project-level conventions provide a form of pre-submission governance by constraining what the agent can do. But convention files are advisory, not enforced. A more robust pre-submission governance layer — sandboxing agent activity, isolating credentials during generation, capturing reasoning traces — appears to be on the roadmap but isn&#8217;t in the August release.</p>



<h2 class="wp-block-heading">Licensing and Availability</h2>



<p class="wp-block-paragraph">Autopilot&#8217;s coding agent capability is included with existing Studio licenses — Enterprise, Community, or trial — subject to a monthly usage quota per user license. This is a significant decision: UiPath didn&#8217;t create a new SKU or premium tier for the coding agent capability.</p>



<p class="wp-block-paragraph">Some important details on the licensing model:</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Item</th><th>Detail</th></tr></thead><tbody><tr><td><strong>Included with</strong></td><td>Studio Enterprise, Community, or Trial license</td></tr><tr><td><strong>Usage model</strong></td><td>Monthly quota per user license</td></tr><tr><td><strong>Unlimited actions</strong></td><td>Expression fixes, commit message generation (don&#8217;t consume quota)</td></tr><tr><td><strong>Quota check</strong></td><td>Run <code>/usage</code> in Autopilot chat</td></tr><tr><td><strong>Top-up</strong></td><td>Administrators can add quota using Platform Units</td></tr><tr><td><strong>Studio Desktop</strong></td><td>GA in STS (2026.0.199+); LTS coming in 2026.10</td></tr><tr><td><strong>Studio Web</strong></td><td>Coming soon</td></tr><tr><td><strong>VS Code</strong></td><td>Pre-release extension available</td></tr></tbody></table></figure>



<p class="wp-block-paragraph">The LTS availability in version 2026.10 is particularly important for enterprise customers who run on UiPath&#8217;s Long Term Support branch rather than the Short Term Support releases. Teams that don&#8217;t adopt STS releases will need to wait until Q4 2026 for the coding agent capability set.</p>



<h2 class="wp-block-heading">What This Changes for RPA Teams</h2>



<p class="wp-block-paragraph">The practical implications differ by role:</p>



<h3 class="wp-block-heading">For RPA Developers</h3>



<p class="wp-block-paragraph">The build-test-debug cycle collapses. Instead of manually dragging activities, configuring properties, running the workflow, checking logs, and iterating, you describe what you want, review what the agent produces, and approve the deployment. Selector debugging — historically one of the most time-intensive tasks — becomes a conversation: &#8220;this selector broke after the application update, fix it.&#8221; The developer&#8217;s value shifts from knowing which activity to use and where to drag it, to knowing what the automation should accomplish and whether the agent&#8217;s output achieves it correctly.</p>



<p class="wp-block-paragraph">This doesn&#8217;t eliminate the need for UiPath expertise. Understanding the platform&#8217;s execution model, Object Repository patterns, Orchestrator queue behavior, and production error signatures remains essential — the difference is that you exercise that knowledge through review and direction rather than manual construction. Developers who invest in understanding the skills architecture and AGENTS.md conventions will have significantly more control over agent output quality than those who rely on generic prompting.</p>



<h3 class="wp-block-heading">For Solution Architects</h3>



<p class="wp-block-paragraph">The ability to plan from PDDs and SDDs changes the solution design workflow. Architects can iterate on design documents knowing that the coding agent will translate them directly into implementations — the feedback loop between design and build tightens from weeks to hours. The AGENTS.md file becomes a critical governance artifact: it&#8217;s where architectural decisions, coding standards, and integration patterns are codified in a format that both humans and agents follow.</p>



<h3 class="wp-block-heading">For Center of Excellence (CoE) Leaders</h3>



<p class="wp-block-paragraph">The democratization pitch is real but requires infrastructure. When &#8220;anyone can describe what they want and direct a coding agent to produce it,&#8221; the CoE&#8217;s role shifts from building automations to governing the platform surface that agents operate on. Skills authoring, AGENTS.md templates, deployment approval workflows, and quota management become the new core competencies. The <a href="https://rpabotsworld.com/rpa-to-agentic-ai-transition-guide/">transition from traditional RPA to agentic workflows</a> accelerates, and CoEs that haven&#8217;t started preparing may find themselves scrambling.</p>



<h3 class="wp-block-heading">For Business Analysts and Process Owners</h3>



<p class="wp-block-paragraph">The barrier to creating automations drops to the ability to describe the process clearly. This sounds simple, but clear process description — the kind that a coding agent can act on — is itself a skill. Teams that have invested in good PDD discipline will extract more value from Autopilot than those whose process documentation is vague or outdated. The PDD, always important in theory, becomes the literal input to the automation engine.</p>



<h2 class="wp-block-heading">Autopilot Coding Agent vs. Traditional Studio Development</h2>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Dimension</th><th>Traditional Studio Development</th><th>Autopilot Coding Agent (Aug 2026)</th></tr></thead><tbody><tr><td><strong>Input</strong></td><td>Manual drag-and-drop, property configuration</td><td>Natural language description, PDD/SDD documents</td></tr><tr><td><strong>Selector building</strong></td><td>UiExplorer + manual tuning</td><td>Agent-generated with Object Repository patterns</td></tr><tr><td><strong>Error handling</strong></td><td>Manually added Try-Catch blocks</td><td>Agent-structured with logging and retry logic</td></tr><tr><td><strong>Debug cycle</strong></td><td>Run → check output → edit → repeat</td><td>Agent runs, observes, diagnoses, proposes fix</td></tr><tr><td><strong>Deployment</strong></td><td>Manual publish to Orchestrator</td><td>Agent chains publish → deploy → run via skills</td></tr><tr><td><strong>Operations</strong></td><td>Orchestrator UI or PowerShell</td><td>Natural-language operational commands</td></tr><tr><td><strong>Extensibility</strong></td><td>Activity packages, NuGet</td><td>Skills (custom + official), MCP servers, AGENTS.md</td></tr><tr><td><strong>Governance</strong></td><td>RBAC, audit trails, credential vaults</td><td>Same + skills-level conventions (pre-submission gap remains)</td></tr><tr><td><strong>Who can build</strong></td><td>Trained RPA developers</td><td>Anyone who can describe a process clearly</td></tr></tbody></table></figure>



<h2 class="wp-block-heading">Strategic Implications: Dines&#8217; Platform Bet</h2>



<p class="wp-block-paragraph">Understanding this release requires understanding the strategic thesis behind it. Across two earnings calls and an investor product strategy session earlier in 2026, Daniel Dines laid out a view of enterprise software economics that explains why UiPath chose to build the platform layer rather than the agent itself.</p>



<p class="wp-block-paragraph">The argument goes like this: as code generation gets cheap — and it&#8217;s getting cheap fast, with frontier models improving quarterly — durable value migrates to the layer that provides trust, integration, and accountability. The coding agent is a commodity that improves with every model release from Anthropic, OpenAI, Google, or open-source alternatives. The orchestration layer is the constant. By positioning as the orchestration platform for <em>every</em> coding agent, UiPath avoids betting on which model wins and instead compounds value with each one.</p>



<p class="wp-block-paragraph">This is a direct counter to the vertical integration approach. Microsoft is building Copilot Studio as a tightly integrated agent-and-orchestration stack within the <a href="https://rpabotsworld.com/microsoft-copilot-studio-august-2026-rebuilt-agent-platform-guide/">Power Platform ecosystem</a>. Automation Anywhere is embedding its own AI capabilities natively. <a href="https://rpabotsworld.com/salesforce-agentforce-multi-agent-orchestration-2026/">Salesforce Agentforce</a> builds agents purpose-specific to the CRM domain. UiPath&#8217;s bet is that enterprises won&#8217;t want to lock into any one model provider — a bet supported by current enterprise buying behavior, where most organizations are hedging across two to three AI vendors.</p>



<p class="wp-block-paragraph">The <a href="https://rpabotsworld.com/uipath-vs-automation-anywhere-vs-blue-prism-agentic-platforms-2026/">competitive landscape among RPA platforms</a> makes this positioning particularly interesting. UiPath holds roughly 35.8% market share and serves 8 out of 10 Fortune 500 firms. By opening its platform to third-party coding agents rather than forcing customers onto a proprietary agent, it&#8217;s betting that the orchestration moat is deeper than the agent moat. If Claude Code is stronger at long-context refactoring this quarter and Codex is stronger at greenfield generation the next, UiPath wins both.</p>



<p class="wp-block-paragraph">The risk, of course, is that the orchestration layer itself gets commoditized — by the <a href="https://rpabotsworld.com/top-trending-open-source-agentic-ai-repos/">rapidly growing open-source ecosystem</a>, by cloud providers bundling orchestration into their agent platforms, or by the coding agents themselves growing capable enough to handle deployment and governance without a separate platform. Dines is betting that enterprise trust requirements — credential isolation, audit trails, regulatory compliance, durable execution — create a switching cost that open-source orchestration tools can&#8217;t easily replicate. For now, with Maestro built on Temporal&#8217;s battle-tested infrastructure, that bet looks defensible.</p>



<h2 class="wp-block-heading">Frequently Asked Questions</h2>



<h3 class="wp-block-heading">Do I need a new license for Autopilot&#8217;s coding agent features?</h3>



<p class="wp-block-paragraph">No. Autopilot as a coding agent is included with your existing Studio license (Enterprise, Community, or Trial). It runs on a monthly usage quota per user license, with certain actions like expression fixes remaining unlimited. Administrators can top up quota with Platform Units if needed.</p>



<h3 class="wp-block-heading">Can I use Claude Code or Codex with UiPath instead of Autopilot?</h3>



<p class="wp-block-paragraph">Yes. UiPath for Coding Agents supports Claude Code and OpenAI Codex today, with additional integrations planned. These third-party agents use the same skills infrastructure as Autopilot. You can run Autopilot in Studio Desktop while other teams use Claude Code or Codex against the same Orchestrator environment — the orchestration and governance layer is agent-agnostic.</p>



<h3 class="wp-block-heading">What&#8217;s the difference between UiPath skills and MCP servers?</h3>



<p class="wp-block-paragraph">Skills are task-oriented instruction bundles that teach agents how to accomplish UiPath-specific tasks using the <code>uip</code> CLI. They can encode multi-step workflows with conditional logic. <a href="https://rpabotsworld.com/mcp-2026-07-28-stateless-spec-agentic-ai-guide/">MCP servers</a> expose tools through a standardized protocol for external service integration. Skills are UiPath&#8217;s primary AI integration path; MCP is supported for specialized setups and is how non-supported agents connect via <code>uip mcp</code>.</p>



<h3 class="wp-block-heading">Will Autopilot as a coding agent be available in Studio Web?</h3>



<p class="wp-block-paragraph">Studio Web support is listed as &#8220;coming soon.&#8221; The coding agent capability set arrives in Studio Desktop LTS with version 2026.10, which is the relevant milestone for enterprise customers running the Long Term Support branch.</p>



<h3 class="wp-block-heading">Does the coding agent handle UI automation selectors?</h3>



<p class="wp-block-paragraph">Yes. Autopilot can build UI automation selectors directly from a prompt, including data extraction workflows. It generates selectors using Object Repository patterns for resilience, and can debug and fix broken selectors — including those that broke after application UI updates.</p>



<h2 class="wp-block-heading">Key Takeaways</h2>



<ul class="wp-block-list">
<li><strong>Autopilot is now a full coding agent</strong> — GA as of August 12, 2026 in Studio Desktop STS (2026.0.199+). It plans from specs, builds workflows, runs them, debugs failures, and restructures code — all through natural language.</li>



<li><strong>Skills-based architecture</strong> is the core design choice. Over 30 built-in tools, an auto-loaded context-aware skills catalog, and the ability to author custom skills give teams precise control over agent behavior.</li>



<li><strong>UiPath for Coding Agents</strong> (launched May 2026) opens the platform to Claude Code, Codex, and future agents. The orchestration layer is vendor-neutral by design.</li>



<li><strong>Operating UiPath via coding agent</strong> also went GA on August 3 — natural-language operations across folders, jobs, queues, assets, audit logs, and more.</li>



<li><strong>Governance covers the output</strong> but not yet the generation process. Pre-submission governance (reasoning traces, credential isolation during build, agent sandboxing) is an acknowledged gap for regulated industries.</li>



<li><strong>No new license required.</strong> The coding agent capability is included with existing Studio licenses, subject to monthly usage quotas.</li>



<li><strong>AGENTS.md and MCP support</strong> mean project-level conventions and external integrations travel with the project regardless of which agent is driving.</li>



<li><strong>The strategic bet:</strong> UiPath is positioning the orchestration layer — not the agent — as the durable value layer in the enterprise AI stack. Whether that bet pays off depends on how quickly coding agents mature and whether the orchestration moat holds.</li>
</ul>



<h2 class="wp-block-heading">References</h2>



<ol class="wp-block-list">
<li>UiPath, &#8220;Autopilot — August 2026 Release Notes,&#8221; August 12, 2026. <a href="https://docs.uipath.com/autopilot/other/latest/release-notes/august-2026" rel="nofollow noopener" target="_blank">docs.uipath.com</a></li>



<li>UiPath, &#8220;UiPath for Coding Agents — August 2026 Release Notes,&#8221; August 3, 2026. <a href="https://docs.uipath.com/coding-agents/standalone/latest/release-notes/august-2026" rel="nofollow noopener" target="_blank">docs.uipath.com</a></li>



<li>UiPath Newsroom, &#8220;UiPath Becomes First Business Orchestration &amp; Automation Platform with Native Integration for Coding Agents,&#8221; May 12, 2026. <a href="https://www.uipath.com/newsroom/uipath-for-coding-agents-launch" rel="nofollow noopener" target="_blank">uipath.com</a></li>



<li>MacQueen, A., &#8220;UiPath opens its platform to every coding agent — here&#8217;s why Claude Code and Codex go first,&#8221; diginomica, May 12, 2026. <a href="https://diginomica.com/uipath-opens-its-platform-every-coding-agent-heres-why-claude-code-and-codex-go-first" rel="nofollow noopener" target="_blank">diginomica.com</a></li>



<li>UiPath, &#8220;Enterprise Automation Platform for Coding Agents,&#8221; 2026. <a href="https://www.uipath.com/developers/coding-agents" rel="nofollow noopener" target="_blank">uipath.com</a></li>



<li>UiPath, &#8220;UiPath CLI — Skills,&#8221; 2026. <a href="https://docs.uipath.com/uipath-cli/standalone/latest/user-guide/concepts-skills" rel="nofollow noopener" target="_blank">docs.uipath.com</a></li>



<li>DevOps Digest, &#8220;UiPath for Coding Agents Released,&#8221; 2026. <a href="https://www.devopsdigest.com/uipath-for-coding-agents-released" rel="nofollow noopener" target="_blank">devopsdigest.com</a></li>



<li>UiPath Community Forum, &#8220;Autopilot is now a coding agent in Studio — generally available,&#8221; August 2026. <a href="https://forum.uipath.com/t/autopilot-is-now-a-coding-agent-in-studio-generally-available/5768853" rel="nofollow noopener" target="_blank">forum.uipath.com</a></li>



<li></li>
</ol>
]]></content:encoded>
					
					<wfw:commentRss>https://rpabotsworld.com/uipath-autopilot-coding-agent-ga-guide-2/feed/</wfw:commentRss>
			<slash:comments>6</slash:comments>
		
		
			</item>
	</channel>
</rss>

<!--
Performance optimized by W3 Total Cache. Learn more: https://www.boldgrid.com/w3-total-cache/?utm_source=w3tc&utm_medium=footer_comment&utm_campaign=free_plugin

Page Caching using Disk: Enhanced 

Served from: rpabotsworld.com @ 2026-11-20 11:29:23 by W3 Total Cache
-->