Microsoft says Foundry now lets organizations build autonomous agents that work in Microsoft 365 under their own identities. Each agent can have its own permissions, manager and audit trail, so its actions are attributed to the agent rather than a person in a conversation.
According to Microsoft, these agents have both a Microsoft Entra agent identity and an Entra agent user account that lets them take Microsoft 365 actions as themselves. An organization can grant an agent access to resources and give it its own email, calendar, OneDrive and Teams presence. Its place in the org chart shows who manages it.
Developers create an agent blueprint that defines capabilities such as the APIs and MCP tools it can use, then deploy it as a Foundry hosted agent. After an admin approves the published blueprint, approved people in the organization can create their own instances. Each instance gets its own identity and access. The blueprint itself does not grant access to a team’s business resources.
Administrators can manage and disable agents
Microsoft says a single control can disable every agent created from a blueprint. It also says administrators can manage agents through Entra and Agent 365. Those controls matter when an agent works across shared spaces such as group chats, where acting with one participant’s permissions could give others access to content they cannot normally see.

The company says its employees have built more than 30 blueprints on Foundry. One agent, called Office of Amanda, is running in production at Microsoft. It uses delegated access to its manager’s calendar and mailbox, and sends email under its own identity.
Nous Research brings Hermes Agent to Microsoft 365
Microsoft also says Nous Research’s Hermes Agent can now work as an AI coworker with its own Microsoft 365 account. Nous plans to offer Hermes Business as a hosted service and Hermes Enterprise for companies to host in their own Azure tenants as Foundry hosted agents. Microsoft says companies will be able to get those offerings soon.



