At Ignite 2025 Microsoft unveiled Agent 365, a new enterprise control plane for managing fleets of AI agents across organizations. The announcement framed Agent 365 as the company’s answer to the growing need for identity, governance, and observability around autonomous agents, promising a single place where IT, security, and developers can register, monitor, and apply policy to agents at scale.
Microsoft Agent 365 is a new enterprise control plane for AI agents is a centralized dashboard and set of services that lets IT teams register, govern, monitor, secure, and measure fleets of autonomous agents running across an organization. It assigns every agent a unique Agent ID, ties agent identities into Microsoft Entra for access control, and surfaces telemetry, dashboards, alerts, and relationship maps so teams can see which agents are talking to which users and data sources.
Agent 365 is designed to treat agents like first-class workforce members, the same way HR and IT manage people, applying Zero Trust principles, role-based and risk-based access, policy guardrails, and integration with Defender and Purview for threat and data governance.
Core capabilities and platform components
Microsoft built Agent 365 around a single, authoritative registry that gives every agent a persistent Agent ID and lifecycle management tied into Microsoft Entra. That registry acts as the system of record for agents, so identity, authentication, and token flows are first-class developer primitives rather than ad-hoc hacks.
Layered on top of identity are granular access controls and policy enforcement: risk-aware permission blueprints that administrators can apply and inherit across agent classes, limiting what each agent can discover, call, or change. Observability is baked in, unified telemetry, dashboards, alerts, and relationship maps make it possible to trace which agents interact with which users and data sources, surface anomalous behavior quickly, and run forensic audits when needed.
Interoperability was a clear design goal: Agent 365 isn’t just for Microsoft-built agents but for third‑party and open‑source agents as well, so teams can bring heterogeneous agent implementations under a single governance fabric. Finally, the platform folds in Microsoft security and governance services, Defender for threat detection and Purview for data classification and policy enforcement, so organizations can combine agent automation with familiar compliance and risk controls.
Availability and how developers and IT can test it
Agent 365 was introduced at Ignite as an enterprise preview available through Microsoft’s early access channels, with initial functionality accessible via the Microsoft 365 Admin Center and related preview programs. For developers, the practical entry points are the Entra SDKs and the specialized OAuth and token flows that let agents authenticate and receive scoped permissions tied to their Agent ID.
IT teams can enable the Agent Dashboard in the admin center to discover active agents and begin mapping risk and data access; from there, pairing Agent 365 with Purview policies and Defender configurations gives a secure sandbox for testing agent behaviors against real-world governance scenarios. In short, the path to trial runs through the standard Microsoft preview/early access route: enable the preview in the tenant admin experience, configure Entra-based identity for your agents, and attach Purview and Defender policies to validate compliance and security during testing.
