The management, governance, and security plane for AI agents in Microsoft 365 - treat agents as first-class identities with the same controls you apply to users.
Overview
Microsoft Agent 365 is the enterprise control plane for AI agents built with Copilot Studio, Azure AI Foundry, and third-party frameworks. It gives agents identities in Microsoft Entra, applies conditional access and RBAC to their actions, meters usage, exposes telemetry, and lets IT govern which data an agent can touch and which apps it can call. As agents move from experiment to production, Agent 365 is the layer that keeps them observable, secure, and compliant.
Key features
Agent identity: Every agent gets a first-class identity in Microsoft Entra - assign entitlements, apply Conditional Access, and audit its actions the same way you would for a user.
Governance and controls: Central policy for what agents can access, which apps they can invoke, and which data they can touch.
Usage metering: Meter and chargeback agent activity, model usage, and downstream API calls across the tenant.
Observability: Unified telemetry and audit logs across all agents, including those from Copilot Studio, Azure AI Foundry, and partner frameworks.
Security integration: Signal flows into the Defender XDR stack for threat detection specific to agentic activity.
Multi-framework: Works with agents built in Microsoft's tooling and with third-party frameworks that adopt the open agent protocols.
Built for
Enterprises rolling out AI agents beyond pilot into production.
IT and security teams that need to apply user-grade controls to agents.
Compliance and finance teams that need auditable agent activity and cost attribution.
Organizations standardizing agent operations across multiple frameworks.
What you get
Annual subscription license for Microsoft Agent 365
Per-tenant management plane assigned via the Microsoft 365 admin center
Access to all Agent 365 features and updates released during the term