Windows 11 Integrates Model Context Protocol to Power AI Agents with Enhanced Security

On May 19, 2025, Microsoft announced that Windows 11 now includes the Model Context Protocol (MCP). Built by Anthropic and open-sourced, MCP lets AI agents talk directly to apps, services, and tools inside the OS. Microsoft says this will help developers build agents that can actually get things done—from managing files to triggering tasks in third-party software.

Microsoft isn’t giving these agents free rein. Users must approve each connection. The OS logs every interaction, and only pre-registered, verified tools appear to agents. Developers can control access through new APIs like App Actions, which limit what agents can trigger in an app.

Also read: Microsoft Build 2025: Top 5 announcements highlights from the developer conference

Microsoft sees MCP as a key building block for its vision of an “agentic OS”—an operating system where AI acts more like an assistant than a chatbot. It’s also working with AMD, Intel, Nvidia, and Qualcomm to make sure this tech runs well on new chips with dedicated AI support.

By locking MCP into Windows, Microsoft is trying to set the standard for how AI agents should work on desktops: useful, but always on a leash. Developers can try it now through the Windows Insider Program.