AI agents now read documents, call APIs, send emails and make decisions on behalf of users. Most organisations were not designed to manage these "non-human actors" at scale. Teams build agents in different tools, security cannot see them all, and nobody can say what each agent can access. Microsoft calls this agent sprawl.
Microsoft Agent 365 is Microsoft's answer: a control plane for AI agents, generally available for commercial customers since 1 May 2026.
Observe, govern, secure
- Observe: one registry of every agent in the organisation, Microsoft and non-Microsoft, including shadow agents discovered in the tenant, plus an agent map and usage insights.
- Govern: lifecycle management, ownership, access reviews and policies through the Microsoft 365 admin center, Microsoft Entra and Microsoft Purview.
- Secure: Entra enforces risk-based access for agents, Purview applies data protection and DLP to agent activity, and Defender detects threats and unsafe behaviour.
Agents get an identity
Through Microsoft Entra Agent ID, each agent becomes a first-class identity. That means the controls you already use for people, such as Conditional Access, access reviews and lifecycle workflows, now apply to agents too.
Which agents does it cover?
Agents built in Microsoft Foundry, Copilot Studio and Agent Builder integrate automatically. Agents on other platforms, such as Amazon Bedrock or Google Vertex AI, can be synchronised into the registry, and custom agents can connect through the Agent 365 SDK.
Licensing
Agent 365 is available as a stand-alone, per-user subscription for eligible Microsoft 365 plans, and it is included in Microsoft 365 E7. Microsoft notes it works best with Microsoft 365 E5 as the foundation.
How Regulus can help
We set up the Agent 365 registry, Entra Agent ID, Purview and Defender controls, onboard your existing agents, and run monthly agent posture reviews as a managed service.