This project provides a Model Context Protocol (MCP) server to enable Retrieval Augmented Generation (RAG) over Apple Notes, allowing AI assistants like Claude to semantically search and reference personal notes. It targets users who want to integrate their Apple Notes knowledge base into AI conversations for enhanced context and recall.

How It Works

The server leverages on-device all-MiniLM-L6-v2 embeddings for semantic search and includes full-text search capabilities. It utilizes LanceDB for vector storage and integrates natively with Apple Notes via JXA (JavaScript for Automation). The architecture is designed for fully local execution, eliminating the need for API keys and ensuring data privacy.

Quick Start & Requirements

• Install: git clone https://github.com/RafalWilinski/mcp-apple-notes followed by bun install.
• Prerequisites: Bun, Claude Desktop.
• Setup: Requires modifying Claude Desktop's claude_desktop_config.json to point to the local MCP server. Indexing is initiated by prompting Claude.
• Docs: https://github.com/RafalWilinski/mcp-apple-notes

Highlighted Details

• On-device embeddings (all-MiniLM-L6-v2) for privacy and local execution.
• LanceDB for efficient vector storage.
• Native Apple Notes integration via JXA.
• MCP-compatible server for AI assistant integration.

Maintenance & Community

The project is maintained by RafalWilinski. There are no explicit community links (Discord/Slack) or roadmap details provided in the README.

Licensing & Compatibility

The repository does not explicitly state a license. This lack of a specified license may imply all rights are reserved or that it is not intended for redistribution or modification. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

Apple Notes are currently returned in HTML format, with a stated intention to convert them to Markdown. Chunking of source content is also a planned feature. The project is still under active development with several items listed in the "Todos" section, indicating potential instability or missing functionality.