Agent 365 launches as Foundry-hosted agent governance, priced at $15/month

AI agents are starting to look less like chatbots and more like employees with permissions, tools, and room to cause trouble. That is the real backdrop here. Microsoft’s news is that Agent 365 went generally available on May 1, 2026, as the company’s control plane for governing, observing, and securing those agents across an enterprise. The pitch is straightforward — if companies are going to let agents act, they need the same kind of visibility and policy controls they already use for people. ### What is Agent 365, really? Basically, it is an admin and security layer for AI agents. Microsoft describes it as a control plane that lets IT and security teams inventory agents, apply access controls, monitor behavior, and manage lifecycle policies across agents built by Microsoft and by third parties. That includes agents coming from Microsoft Foundry, Copilot Studio, and other sources that get registered or discovered in the tenant. ### What changed on May 1? The big change is availability. Agent 365 had been previewed earlier, but on May 1 Microsoft moved it into general availability for commercial customers. Microsoft announced the same date for Microsoft 365 E7, the new suite that bundles Copilot, Microsoft 365 E5, Entra Suite, and Agent 365 together. So this is not a concept demo anymore — it is a product Microsoft wants enterprises to buy and roll out now. ### How is it priced? The headline number is $15 per user per month as a standalone license. Microsoft also includes it in Microsoft 365 E7. The licensing logic is interesting — one license covers a person who manages or sponsors agents, or who uses agents on their behalf. So Microsoft is pricing governance around the human owner or operator, not charging per individual agent action. ### Why does Foundry matter here? Foundry is Microsoft’s developer platform for building AI apps and agents, so tying Agent 365 into Foundry means governance is getting pushed closer to where agents are created and deployed. Microsoft’s documentation frames Agent 365 as the way to apply identity, security, governance, and lifecycle control ### What security pieces are actually in the stack? This is where the product gets more concrete. Agent 365 extends Microsoft’s existing security stack — especially Entra, Defender, and Purview — to agents. Purview handles data security and compliance controls for supported agent scenarios. Microsoft also says Agent 365 adds runtime protection and broader observability for agents operating with their enterprises. ### Is this about Microsoft-only agents? Not entirely — and that is part of the strategy. Microsoft keeps stressing that Agent 365 is meant for heterogeneous environments, not just Microsoft-built agents. The company is trying to make governance the sticky layer above the model and framework wars. That matters because enterprises increasingly want Claude, OpenAI models, and Microsoft tools in the same environment without building separate policy systems for each one. ### Why launch this now? Because “agent sprawl” is becoming the next version of SaaS sprawl. Once teams can create agents quickly, companies lose track of what exists, what data those agents can access, and what actions they can take. Microsoft is betting that the winner in enterprise AI will not just offer the best models or builders — it will offer the safest operating system for a messy fleet of semi-autonomous software workers. ### Bottom line Agent 365 is Microsoft formalizing a new category: agent governance as core enterprise infrastructure. The product itself is not flashy. But that is the point — Microsoft wants AI agents to plug into the same admin, identity, security, and compliance machinery that already runs big companies. If agents really are becoming digital coworkers, Agent 365 is Microsoft’s argument that they need managers too.