Local MCP: Free macOS Tool Gives Claude, ChatGPT Direct Access to Local Files and Apps
Key Takeaways
- ▸Local MCP connects AI assistants to 180+ local Mac apps while keeping data on-device by default, with an optional encrypted relay for web clients
- ▸The tool reads message history and local app data that standard cloud APIs cannot access by directly reading local data stores like IndexedDB and LevelDB
- ▸Hundreds of users are actively running it with thousands of daily operations; free to install and use with no paid tier planned
Summary
Local MCP, a native macOS application, connects AI assistants like Claude, ChatGPT, and Cursor to local applications and personal files while keeping all data on-device by default. The tool provides approximately 180 integrations with Mac apps including Mail, Calendar, iMessage, Teams, Slack, WhatsApp, OneDrive, Google Drive, Office documents, and local files. For desktop clients (Claude Desktop, Cursor, Claude Code), everything runs on localhost with no data leaving the machine; for web-based AI assistants that cannot access localhost, users can optionally enable an encrypted relay that is off by default.
Developed by lanchuske, Local MCP solves a fundamental limitation of cloud-based AI assistants: they lack access to personal context that lives only on user machines. The tool reads data directly from native apps via system APIs (EventKit for Calendar/Contacts, AppleScript/JXA for Mail/Messages) and by accessing local data stores (IndexedDB/LevelDB) for chat apps like Teams and Slack — allowing access to message history that cloud APIs cannot reach. Destructive actions like sending emails or deleting events always require explicit user preview and confirmation.
The project has moved past the experimental stage with hundreds of active users running thousands of operations daily. It is free with no planned paid tier, signed and notarized for macOS 13+, and available immediately for download or installation via command line. The architecture is GDPR compliant by design with no API keys, tokens, or OAuth flows required.
- Privacy-first design: GDPR compliant, zero API keys or OAuth, all processing on-device for desktop clients (Claude Desktop, Cursor, Claude Code)
Editorial Opinion
Local MCP elegantly solves a genuine pain point: AI assistants are most useful when they understand your context—your emails, calendar, messages, files—yet cloud models fundamentally cannot access personal data without risky API key sharing or complex OAuth flows. By localizing 180+ tool integrations and making the data-relay opt-in rather than default, lanchuske has built something that actually respects privacy while making AI assistants more capable for real work. The fact that hundreds are already using it daily at scale suggests this fills a critical gap in the AI assistant ecosystem.
