On June 3, 2026, Microsoft introduced new features for its Azure AI Foundry platform, designed for the development and management of enterprise AI agents. The update enables the direct publication of agents into Microsoft 365 Copilot and Teams, streamlining deployment for organizational workflows.
The platform now includes public preview access to Autopilot agents and Agent-to-Agent functionality, allowing coordination between multiple agents. Every agent created is assigned an identity via Microsoft Entra ID and registered in Microsoft Agent 365 for centralized oversight by IT and compliance teams.
Microsoft plans to reach general availability for model deployment policies by August 2026, allowing companies to restrict model usage based on specific criteria. Additionally, a new Foundry Control Plane is under development to provide comprehensive fleet-wide management for all deployed agents.
Starting in September 2026, long-running agents will receive resilience updates, including durable identities and lease-based recovery. The platform also supports per-session isolation, persistent file systems, and scale-to-zero economics to optimize resource consumption and security for enterprise tasks.
