Microsoft Agent 365 hits GA: $15/user/month for the agent control plane

Microsoft shipped Agent 365 to general availability on May 1, 2026 — a dedicated governance and security control plane for enterprise AI agents. The Microsoft Security Blog frames it as “the control plane for agents”: a single layer for IT, security, and compliance teams to observe, govern, and secure AI agents — across vendors, not just Microsoft’s own.

What Agent 365 actually is

Three product pillars:

• Observe — discover AI agents running across an organization (sanctioned and shadow), map their permissions, see what tools they’re calling
• Govern — apply policies for which agents can run, what data they can access, what actions they can take, who can approve high-risk operations
• Secure — runtime blocking, alerts on policy violations, audit logs, integration with Microsoft Defender and Intune

Crucially, Agent 365 is agent-vendor-agnostic. It governs ChatGPT-based agents, Claude-based agents, Copilot, internal custom agents, and third-party tools all in one pane. Microsoft is positioning itself as the control layer regardless of which AI vendor enterprises actually buy.

Pricing

• Agent 365 standalone: $15/user/month
• Microsoft 365 E7 (new SKU): $99/user/month — bundles E5 ($60), Copilot ($30), Agent 365 ($15), and Entra Suite ($12). Effectively zero-margin on the bundle if you’d buy each separately.
• Public preview: context mapping, policy-based controls, runtime blocking, and Defender/Intune integration in June 2026

The $99 E7 bundle math is interesting: it makes Copilot + Agent 365 + Entra “free” if you’d already considered E5 + Copilot, which most enterprise Microsoft shops are.

Why this matters

The narrative shift is what’s noteworthy. For two years, the enterprise AI conversation was “which chatbot do we deploy?” Then “which coding assistant?” Then “how do we measure agent quality?” Microsoft’s Agent 365 launch reframes 2026’s question as “how do we govern agent fleets across vendors?”

That reframe favors Microsoft for two reasons. First, it elevates the buyer from a developer-tools manager to the CISO — a buyer who already has Microsoft contracts. Second, it makes vendor lock-in into agents (Anthropic, OpenAI, Google) less of a strategic risk for the enterprise, because the governance layer is consistent regardless of which agent runtime is underneath.

What it means for the broader market

For GitHub Copilot: Agent 365 strengthens Copilot’s enterprise position. The combined “Copilot writes the code, Agent 365 governs the agents that ship it” pitch is structurally hard for Cursor, Claude Code, or independent agent tools to match.

For OpenAI Operator, Manus AI, agent platforms generally: if Agent 365 becomes the de facto enterprise governance layer, these tools become “agents that need to integrate with Agent 365” rather than full-stack solutions. That’s a healthy ecosystem dynamic — and a real positioning constraint.

For Anthropic’s Pentagon shutout: the Pentagon’s May 1 deals with Microsoft, OpenAI, and others (excluding Anthropic over safety guardrail disagreements) overlap with the Agent 365 launch by hours. The framing — “Microsoft can govern AI agents at federal scale” — is consistent with what the Pentagon was buying.

The pragmatic read

Microsoft Agent 365 is the most significant enterprise AI infrastructure launch of 2026 so far. It’s not a model. It’s not a chatbot. It’s the layer underneath all of those, sold to a buyer (the CISO) who has different incentives from the developer-tools buyer who picks Cursor or Claude Code.

For solo developers and indie users — Pick Right’s main audience — Agent 365 is irrelevant. For anyone working at a company that has procurement: expect to see Agent 365 in 2026 IT roadmap discussions.

For more on enterprise AI agents, see best AI agents in 2026.

Questions or corrections? Email Pick Right. Want the full list? See all news.