Hunch uses MCP to automate macOS tasks via local AI

The open-source project Hunch utilizes the Model Context Protocol (MCP) to enable local language models to automate tasks directly within macOS. By employing a local protocol server, the architecture allows AI models to act as coordinators that translate user intentions into specific system calls and actions.

A primary feature of this approach is the emphasis on privacy and security. Because the processing occurs locally on the Mac, sensitive data, system configurations, and file contents are not transmitted to external cloud servers. This setup allows desktop applications, such as Claude Desktop, to interact with the operating system through a standardized interface.

The system is designed to handle various operating system functions, including file system operations like directory browsing and folder creation, as well as opening applications. This decoupling of AI logic from operating system logic provides a scalable way to integrate automation without manual wiring for every individual scenario.