Microsoft Agent 365 moved from public preview to general availability on May 1, 2026. For organizations already deep in AI agent deployments, this changes how IT teams actually govern what those agents do.

The platform has been in preview since late 2025, but the GA release brings several features that make it practically useful rather than aspirationally useful. The headline addition is management of locally running AI agents — tools like OpenClaw, GitHub Copilot CLI, and Claude Code that your developers are already using on their Windows laptops. IT admins can now apply Microsoft Intune policies to these tools and use Microsoft Defender to monitor their behavior in real time.

That matters because local AI coding agents have been operating largely outside IT visibility in most organizations. They run on endpoints, they have access to codebases and credentials, and until now there was no standard way to enforce guardrails on them at the organizational level.

What Agent 365 Actually Does

The platform is designed to give enterprises one place to see and control all the AI agents operating across their environment. That includes agents built in Microsoft Copilot Studio, locally installed coding agents on Windows, and now — in public preview — agents running on AWS Bedrock and Google Cloud.

The multicloud registry sync means IT can import AWS Bedrock and Google Gemini Enterprise agents into a single Agent 365 inventory. Full lifecycle governance across those platforms, including the ability to start, stop, and delete agents remotely, is scheduled for June 2026.

Other GA features include:

• A shadow AI dashboard in the Microsoft 365 admin center showing all agents across the organization, including ones IT did not officially deploy
• Network controls extending Microsoft Entra capabilities to Copilot Studio agents and local agents
• Runtime blocking that can automatically halt coding agents exhibiting malicious patterns, like attempting to exfiltrate credentials or make unexpected external network calls
• Policy-based guardrails that set behavioral boundaries for agents before they run

Pricing is $15 per user per month as a standalone product, or included at no additional cost in Microsoft 365 E7 subscriptions.

The Ecosystem That Comes With It

Microsoft announced integration partnerships with Genspark, Zensai, Egnyte, Zendesk, Kasisto, Kore, and n8n. These integrations extend governance visibility into third-party agents built on those platforms, which is where a lot of enterprise agent sprawl actually lives.

The Microsoft 365 E7 inclusion is the nudge that will likely accelerate adoption. Organizations already paying for the top-tier Microsoft 365 bundle now have enterprise-grade agent governance without a separate line item.

What This Means for Business

If you have more than 50 people and you are deploying AI agents, you almost certainly have an inventory problem. Agents are being built by your developers, bought as SaaS tools, installed on personal laptops, and wired into business systems by people who are not waiting for IT approval. That is not a criticism of those people. It is just how fast the category is moving.

The risk is not that agents are dangerous. The risk is that nobody knows what they are doing. An agent with access to a company email account and a CRM system can send messages, update records, and query customer data without a human in the loop. When that goes wrong — and sometimes it does — organizations need to know which agent did what, when, and why.

Agent 365 is Microsoft’s answer to that problem. It is not perfect. The multicloud support is still in preview, and governing agents from every vendor is a hard problem that a single platform will not fully solve in year one. But having a centralized inventory and policy layer is the first step that organizations have been missing.

For teams running Microsoft 365 already, the E7 inclusion means there is now no reason not to start inventorying your agents. The shadow AI dashboard alone is worth enabling just to understand what is actually running in your environment.

At Enterprise DNA, we have been saying for a while that the challenge with AI agents is not building them — it is governing them at scale. Microsoft Agent 365 is a significant step toward making that governable, and the GA release signals that enterprise AI governance is now a standard expectation rather than an advanced capability.

If you want the playbook other teams are using with Claude and Codex right now, grab the free Working With Claude field guide. Download it here.