By Alex Pozin, Samer Baroudi, Brendan Powers
As AI agents become a core part of enterprise operations, organizations need a consistent way to observe, manage, govern, and secure agents across an increasingly distributed ecosystem spanning cloud, local, and partner-developed experiences. Since the GA release, Agent 365 has continued to evolve as the control plane for enterprise agents. This month’s innovations strengthen visibility and protection for local agents, simplify onboarding and lifecycle management, and expand programmatic governance capabilities, helping organizations scale AI adoption with greater confidence.
Visibility, controls and protections for local agents
As local agents become a growing part of the enterprise AI landscape, Agent 365 is expanding end-to-end visibility, governance, and protection across these experiences. From discovering and inventorying local agents in the Agent 365 Registry, to auditing agent activity with Microsoft Purview, securing runtime behavior with Microsoft Defender, and enforcing isolated execution through Microsoft Execution Containers (MXC), organizations gain the controls needed to confidently manage, secure, and govern local agents at scale without sacrificing user productivity.
Local agents in the Agent 365 registry
Organizations increasingly rely on local agents running directly on user devices, yet these agents often operate outside AI visibility. With local agents now in the Agent 365 registry (public preview), IT and security teams gain a unified view of shadow local agents detected across the environment. The registry surfaces each agent's publisher, detected device count, and last scanned date, and opens into a per agent details pane that lists every endpoint where the agent has been observed, complete with device name, model, and operating system. By centralizing discovery and device‑level context for local agents across managed endpoints, teams pinpoint which shadow AI agents users are running, scope their footprint by device, and accelerate investigations.
Microsoft Purview Audit coverage for GitHub Copilot CLI, Claude Code, OpenAI Codex, and OpenClaw
Microsoft Purview Audit in Agent 365 continues to expand its coverage, bringing deeper visibility into agent interactions and strengthening enterprise compliance capabilities. Building on its ability to capture and monitor audit logs across human-to-agent, agent-to-agent, and agent-to-tool interactions with rich contextual metadata on access, actions, and activity, the June update introduces support for emerging local and developer-focused agents, including GitHub Copilot CLI, Claude Code, OpenAI Codex, and OpenClaw. This expansion ensures that organizations can audit activity consistently across these local agents, closing visibility gaps as agent usage becomes more distributed. By extending familiar Microsoft Purview auditing workflows to these new agent experiences, enterprises can more effectively investigate incidents, meet compliance requirements, and confidently scale AI adoption with comprehensive, end-to-end observability.
Microsoft Defender and Purview expand Agent 365 security capabilities to cover local agents (Public Preview)
Microsoft Defender in Agent 365 expands security posture management, on-device runtime protection, XDR alerting, and advanced hunting for local AI agents. Since the June 2 release, Defender has expanded discovery of local agents and MCP servers across managed Windows and macOS devices, now supporting more than 35 known agent types in preview. With this release, Defender also introduces exposure risk for local agents, helping security teams identify and prioritize the riskiest local agents based on their configuration, permissions, and exposure, with recommendations to reduce risk.
Defender's on-device runtime protection capabilities, now available in public preview, support Claude Code, GitHub Copilot CLI, GitHub Copilot app, Codex CLI, OpenClaw, and similar Node-based claws using agent-native event inspection and an on-device, network-based runtime inspection. Defender can audit or block malicious instructions introduced through prompt injection attempts and generates Microsoft Defender XDR alerts when attacks are detected. Integration with Microsoft Purview extends protection by auditing or blocking prompts and tool calls in real-time to help prevent exfiltration of sensitive data based on Purview policies. Advanced hunting enables security teams to investigate local agent activity, hunt for exposure risks, and correlate agent inventory with existing endpoint telemetry.
Microsoft Execution Containers for local agents (Public Preview)
As organizations adopt local AI agents running directly on user devices, they introduce new challenges around visibility, control, and risk, these agents often operate outside traditional governance boundaries, can access sensitive data or tools, and may execute in unmanaged environments. Microsoft Execution Container (MXC) addresses this challenge by enforcing a secure, isolated runtime for local agents, bringing them back under centralized administrative control. In public preview, MXC enables IT to require that local agents run within a hardened Windows container, reducing exposure from unmanaged execution while standardizing security protections across the fleet. By combining isolation with centralized policy enforcement, MXC gives organizations greater confidence in local agent usage, improving visibility into runtime behavior, enabling consistent governance, and mitigating the risk of data leakage or malicious activity without disrupting user productivity.
New Capabilities for Onboarding Agents and Managing the Enterprise Agent Fleet
Organizations need a faster path to enterprise readiness and stronger controls for ongoing governance. Agent 365 advances both goals by simplifying agent onboarding with Agent 365 skills, expanding programmatic management through Microsoft Graph APIs, and extending Microsoft Entra ID Governance to ecosystem partner agents, helping IT manage agents with greater automation, visibility, and control at enterprise scale.
Accelerate agent onboarding for builders with Agent 365 skills
With Agent 365 skills, agent developers can dramatically simplify how they bring agents into Microsoft Agent 365. As outlined in the blog post, this innovation removes the friction of the traditional onboarding “last mile” where developers previously had to manually configure identity, observability, governance, and secure access to Microsoft 365 data. With Agent 365 skills, developers can now use natural language directly within their preferred coding environments to request outcomes like adding observability or integrating tools, while the platform automatically applies and validates the required configurations. By transforming a complex, multi-step setup into a guided, outcome-driven experience, Agent 365 Skills accelerates time to enterprise readiness and enables organizations to operationalize AI agents at scale, faster and with greater confidence.
Expanded Microsoft Graph API support
Agent 365 continues to expand its Microsoft Graph API surface area, giving AI administrators powerful, programmatic control over the agent ecosystem. As outlined in the Agent 365 Graph API documentation, these APIs enable scalable, automated management of agents, moving beyond manual or UI-driven workflows to support enterprise-grade governance, reporting, and lifecycle automation.
At a high level, the Agent 365 Graph APIs allow administrators to:
- Retrieve a complete inventory of agents across the organization
- Access detailed metadata for auditing, compliance, and lifecycle management
- Automate onboarding, governance workflows, and operational processes at scale
- Perform management actions such as blocking, unblocking, or reassigning ownership of agents
These capabilities help IT teams integrate agent management into existing toolchains and workflows, improving visibility, consistency, and control across growing agent estates.
As part of this release, two foundational package management APIs: List packages and Get package details have now moved from beta to Microsoft Graph v1.0, marking an important step toward production-ready automation. Note that these package management APIs return agents. Apps and add-ins are not included in the API response. These APIs enable AI admin to:
- List packages (GET /copilot/admin/catalog/packages): Retrieve a full inventory of all agents in the organization for governance and reporting
- Get package details (GET /copilot/admin/catalog/packages/{id}): Access rich metadata for a specific agent, including configuration, ownership, and deployment details
In addition, we’ve introduced couple of key enhancements driven directly by customer feedback:
- Expanded role-based access: Support for Global Reader and AI Reader roles enables read-only visibility into these APIs without requiring elevated admin privileges, improving delegation and operational flexibility.
- Application permissions support: Customers can now call these APIs using app-only permissions, unlocking secure, unattended automation scenarios.
By bringing these APIs to Graph API V1.0, Agent 365 provides a more stable, supported foundation for building custom automation solutions such as inventory reporting, compliance checks, and lifecycle management workflows, directly on top of Microsoft Graph. This advancement reinforces Agent 365’s role as the control plane for enterprise agent governance, enabling administrators to manage agents with the same rigor, scale, and automation used for users and applications. Learn more about the Agent Package Management API here.
Microsoft Entra ID Governance for ecosystem partner agents
Microsoft Entra ID Governance for agent entitlements brings proven identity governance controls to the AI workforce, enabling organizations to manage agent access with the same rigor as human users through scoped permissions, ownership accountability, and time-bound access. With the June update, support now extends to ecosystem partner agents integrated through the Agent 365 SDK, allowing enterprises to apply entitlement management consistently across both internally-developed and third-party agents. This ensures that these agents are governed through access packages, lifecycle workflows, and least-privilege principles, reducing the risk of over-provisioned agents. This delivers unified access governance at scale, helping enforce policy, maintain compliance, and provide clear ownership and accountability as agent ecosystems expand across internal and partner-built agents.
Conclusion
The June 2026 updates continue to advance Agent 365 as the control plane for enterprise agents. By expanding visibility and security for local agents, streamlining enterprise onboarding, and enhancing automation and governance capabilities, Agent 365 helps organizations scale AI adoption with greater confidence, control, and operational excellence.
Next steps:
Microsoft