There is now a date on the calendar. On May 1, 2026, Microsoft launches Agent 365, a dedicated control plane for managing AI agents across the enterprise. Whether those agents were built by Microsoft, Anthropic, OpenAI, SAP, ServiceNow, or an internal team using LangChain or CrewAI, Agent 365 is designed to govern all of them from a single place.

This is a different product from Copilot. It is infrastructure.

The Problem It Solves

Anyone can spin up an AI agent today. That is, increasingly, the problem. Agents proliferate faster than the security around them. They hold credentials, access sensitive data, and take actions at machine speed. Most organizations have no central view of which agents exist, what they can do, or who authorized them.

Microsoft says that in just two months of preview, tens of millions of agents appeared in the Agent 365 registry. Tens of millions. Most enterprises had no idea that many agents were active inside their environments.

Agent 365 addresses this directly with five capabilities: a registry, access controls, visualization, interoperability, and security. Every agent gets a Microsoft Entra Agent ID, which ties it to the organization’s identity layer. IT teams can inventory known agents and detect shadow agents. Security teams can quarantine unauthorized agents instantly. Compliance teams get audit trails for everything an agent touches.

Not Just for Microsoft Agents

The scope is broader than the Microsoft product ecosystem. Agent 365 supports agents from Anthropic, OpenAI, Cursor, Perplexity, Vercel, and open-source frameworks like LangGraph and CrewAI. Agents from SAP, ServiceNow, Workday, Databricks, Adobe, and NVIDIA are also supported.

The practical meaning: if your organization deploys AI agents across multiple platforms and vendors — which is already the case for most mid-size and large businesses — Agent 365 becomes the single pane of glass for managing them. This eliminates the scenario where finance deploys an accounts payable agent, IT deploys a support ticket agent, and marketing deploys a campaign optimizer, and nobody has a clear picture of what any of them are doing.

The Pricing and Packaging

Agent 365 launches at $15 per user per month as a standalone product. It is also included in Microsoft 365 E7, a new bundle priced at $99 per user per month that packages in M365 E5, Microsoft Copilot, Agent 365, and the Microsoft Entra Suite. Microsoft’s license model covers all agents acting on behalf of a licensed human user, so individual agents do not each require separate licenses.

The E7 tier signals how Microsoft views this moment: AI agent governance is not optional enterprise tooling, it is part of the core security and productivity stack.

What This Means for Business

The businesses watching this should take note of two things.

First, the scale signal. Tens of millions of agents registered in two months of preview means agent deployment has already crossed into the mainstream of enterprise technology, not just the early adopter fringe. Most organizations are already running more agents than they realize. The governance question is not hypothetical.

Second, the architecture signal. Microsoft is positioning identity as the control layer for AI agents, not the application layer. The agent’s permissions, access scope, and lifecycle events flow through the same identity infrastructure that governs human users. This is the right architectural answer. Security teams who have spent years building least-privilege access controls for humans can extend those controls to agents using tools they already understand.

For business owners deploying AI agents outside the Microsoft ecosystem, the underlying principle still applies: every agent needs an identity, defined permissions, and a kill switch. Whether that governance comes through Agent 365, Okta for AI Agents, or a custom-built layer, it needs to exist before the agents go live.

The organizations that build governance infrastructure now will scale faster and with fewer preventable incidents than those who add it after something goes wrong.

Want the practical version of this? The free Working With Claude field guide covers the full Claude ecosystem, Claude Code, and how to roll it out across a real business. Download it here.