Microsoft Unveils Agent 365: A Control Plane for the AI Revolution
Microsoft is addressing a critical challenge in the burgeoning age of artificial intelligence – managing the growing fleet of AI agents – with the introduction of Microsoft Agent 365. Announced at Ignite 2025, the solution aims to provide organizations with unprecedented openness and control over AI agents, regardless of their origin, as these digital workers increasingly integrate into productive processes.
As AI agents find their way into more workflows, ensuring both safety and responsible implementation become paramount. Agent 365 is designed to deliver transparency regarding AI agent observability, extending across all agents utilized within an institution. Crucially, this includes agents developed on Microsoft platforms, and also those originating from open-source frameworks or third-party providers. The solution equips IT teams with the tools needed to securely manage these agents thru telemetry, comprehensive dashboards, and proactive alerts, ultimately reducing risk and closing perception gaps.
The platform functions as a central control level, allowing administrators to map individual agent functions via existing Microsoft security and compliance tools like Microsoft Entra, Defender, and Purview. Agent 365’s capabilities are structured around five core areas: registry, access control, visualization, security, and interoperability.
Establishing an AI Agent Registry and Identity Management
at Microsoft Build 2025,the company also introduced Microsoft Entra Agent ID,a system for managing and securing the identities of AI agents. Building on this, Entra Registry will serve as a unified directory for all agents employed within a company, including those managed with Entra Agent ID. This centralized registry will empower administrators to quarantine unauthorized agents, preventing them from being used by personnel or connecting to critical company resources. The goal is to proactively prevent the uncontrolled proliferation of agents within an organization’s environment.
Granular Access Control for AI Agents
controlling what AI agents can do is as important as knowing what they are. With Entra ID providing a unique identifier for each agent, administrators can enforce resource access permissions mirroring those applied to human users. This allows for the implementation of the principle of least privilege – frequently enough referred to as OCTOPUS – limiting agents to only the resources necessary to complete their assigned tasks. “IT teams can enforce limitations from the outset through policies and templates,” according to Microsoft, enabling context- and risk-based guidelines.
Monitoring Agent Activity and Performance
Beyond defining permissible actions, Agent 365 provides tools for monitoring agent behavior. The Microsoft Agent 365 dashboard offers IT teams a visual depiction of connections between agents, users, and organizational resources. Detailed logging, report creation, and e-revelation functionality are also included to support compliance requirements and detailed investigations.
Perhaps most interestingly, Agent 365 will classify agent performance based on speed, efficiency, and quality. this data will be valuable for decision-makers evaluating the return on investment of AI agents, and also for end-users assessing the effectiveness of assigned tasks.
Security as a Foundational Pillar
Security is not an afterthought but a core component of Agent 365. The solution is designed to protect against both external and internal threats, leveraging existing Microsoft security solutions like Defender, Entra ID, and Purview. Integrating with Microsoft Defender allows for the detection and response to threats targeting agents directly, while Entra ID facilitates prompt blocking of attacks. Microsoft Purview users will be able to identify instances where agents are mishandling data, triggering adaptive policies to mitigate risk and ensure compliance.
Ensuring Seamless Interoperability
Microsoft recognizes that a successful AI strategy requires interoperability. agent 365 is designed to work with agents from various Microsoft environments – including Copilot Studio, the Microsoft Agent Framework, and the Agent 365 SDK – and also open-source frameworks. Moreover, agents must seamlessly access the same applications as human users, frequently enough within the same context, to ensure smooth process execution.
Accessible Through the Microsoft 365 Admin Center
Microsoft Agent 365 is primarily accessed through the familiar Microsoft 365 Admin Center. Participants in Microsoft’s Frontier program will receive early access to new AI functionalities through an exclusive early access program.administrators already utilizing Microsoft Entra, Purview, or Defender will find the corresponding agent 365 functionalities integrated directly within those existing tools.
