Microsoft Launches Agent 365: A Governance Layer for Every AI Agent in Your Enterprise

Microsoft’s Agent 365 is generally available starting today, shipped as part of the new Microsoft 365 E7 enterprise tier. It’s not a coding tool. It’s the layer that sits above your AI coding tools and everything else, governing how agents operate, what they can access, and who can deploy them.

The timing reflects where enterprise AI adoption actually is. A recent Dynatrace survey found that 72% of enterprises are still stuck in AI pilot phases, with security and compliance concerns cited as the primary blockers. Agent 365 is Microsoft’s direct answer to that problem.

What Agent 365 Does

The core function is treating AI agents as managed digital workers rather than ad-hoc tools. Every agent, whether built with Copilot Studio, Microsoft Foundry, the GitHub Copilot coding agent, or a third-party platform, can be registered in a central inventory.

From there, IT and security teams get:

Observability. A complete registry of all agents in the organization with performance and usage analytics. You can see what agents exist, what they’re doing, and whether they’re meeting their intended purpose.

Governance. IT-controlled onboarding workflows, security policy templates, and lifecycle management rules. You can set policies to automatically expire inactive agents, flag ownerless agents, and block agents that don’t meet your security requirements before they touch production systems.

Security. Integration with Microsoft Defender, Entra, and Purview. That means agent identities flow through the same access management infrastructure as human users, data compliance extends to agent outputs, and threat detection applies to agent actions.

Least-privilege access enforcement. You can define exactly which users, data sources, and tools each agent can reach. This matters for coding agents specifically: a Copilot agent scoped to read-only access to a test repo is less risky than one with write access to production.

Pricing

Agent 365 ships as part of Microsoft 365 E7, the new enterprise tier priced at $99 per user per month. E7 bundles E5, Microsoft 365 Copilot, Entra Suite, and Agent 365 together. It’s Microsoft’s first new enterprise license tier since E5 launched in 2015.

Agent 365 is also available as a standalone add-on for organizations that don’t want to move to E7.

Why This Matters for Coding Tools

Most AI coding tool discussions focus on what the model can build. Agent 365 addresses what the enterprise can actually deploy safely.

GitHub Copilot’s coding agent, Codex running in multi-agent configurations, and any custom agents built on Copilot Studio all become manageable through a single governance interface. For enterprise IT teams that have been watching developer adoption of these tools with a mix of interest and alarm, Agent 365 gives them visibility and control without forcing them to block the tools entirely.

The coding agent specifically gets a few relevant capabilities from this. You can scope which repositories and tools the Copilot coding agent can access. You get audit logs of what the agent actually did during an autonomous session. And you can enforce that agents are registered and approved before they can act on production infrastructure.

None of this changes how developers interact with their coding tools day-to-day. It changes how much the enterprise can trust that those tools are operating within policy.

The Agent 365 documentation and pricing details are live on Microsoft.com.

---

Sources: Microsoft Agent 365 overview, Microsoft Agent 365 product page, Trustmarque: M365 E7 and Agent 365 explained, Dynatrace Pulse of Agentic AI 2026