What it does

Agent 365 is Microsoft's control plane for AI agents — the layer for registering, securing, governing, and observing your whole agent fleet, whether the agents were built on Microsoft platforms, open-source frameworks, or third-party tools.

Key facts

• Five capabilities: Registry, Access Control, Visualization, Interoperability, and Security.
• Agents get identities via Microsoft Entra Agent ID, which also surfaces shadow agents.
• Managed from the Microsoft 365 admin centre, with Entra, Purview, and Defender in the loop.
• Supports Microsoft-native (Copilot Studio, Foundry, prebuilt) plus partner and open-source agents.
• Announced at Ignite 2025; initially via Microsoft Frontier early-access.

When to use / skip

Relevant once you have more than a handful of agents, or agents from mixed sources, and need inventory, identity, and oversight. Skip the fleet-management framing for a single pilot agent — the overhead isn't warranted yet.

Configuration decisions

• Whether to enrol through Frontier early access now or wait for GA.
• Which agents to bring under management first (start with the sprawl risk).
• How Entra Agent ID, Purview, and Defender policies apply to agents.

Gotchas

• It's early-access and moving fast — capabilities and licensing will shift, so don't hard-commit a design around today's state.
• "Shadow agents" implies you already have ungoverned agents; discovery may surface more than leadership expects.
• Identity for agents is a new governance object — existing RBAC assumptions don't map cleanly.

Consultant notes

• Position Agent 365 as the governance answer to agent sprawl — it's the story to tell IT leaders nervous about ungoverned agents.
• Licensing and packaging are unsettled post-Ignite; keep proposals conditional until GA terms firm up.
• Treat agent identity as a first-class Entra workstream, not an afterthought bolted onto the agent build.

---

Review after Agent 365 reaches GA and licensing is published.