Microsoft Highlights MXC SDK for Controlled AI Agent Execution on Windows

Microsoft introduced an early preview of the Microsoft Execution Containers (MXC) SDK to provide policy-driven, isolated execution environments for local AI agents on Windows. The SDK allows AI agents to execute code, invoke tools, and complete complex workflows locally while restricting direct access to sensitive user data. As AI agents gain autonomy to run local code and call system tools, isolating their execution is crucial to preventing data leaks and security breaches. MXC serves as a key pillar in Microsoft's ecosystem strategy, connecting local agent activities into enterprise management planes like Agent 365 and Microsoft Security controls. The MXC release features a TypeScript SDK offering one-shot and state-aware APIs alongside Event Tracing for Windows (ETW) diagnostics. However, Microsoft cautions in its open-source repository that current MXC profiles are an early preview and should not yet be relied upon as a strict security boundary.