makenotion/notion-mcp-server

What it solves

This project provides a way for AI agents to interact directly with Notion workspaces. It bridges the gap between LLM-powered clients (like Claude Desktop, Cursor, or Zed) and the Notion API, allowing AI to read, search, and edit Notion pages and databases using a standardized protocol.

How it works

It implements a Model Context Protocol (MCP) server that wraps the Notion API. The server exposes a set of tools (currently 22) that AI agents can discover and call. It supports two communication methods: standard input/output (STDIO) for local desktop clients and a streamable HTTP transport for web-based applications. To optimize for AI, it includes specialized tools that handle page content as Markdown rather than complex JSON blocks, reducing token usage.

Who it’s for

• AI Agent Users: People using MCP-compatible clients who want their AI to manage their Notion notes and databases.
• Developers: Those building AI applications that need a structured, token-efficient way to integrate Notion data.

Highlights

• Token-Efficient Markdown: Specialized tools for reading and updating page content in Markdown format to save on LLM tokens.
• Flexible Deployment: Can be run via npm, Docker, or as a remote server.
• Multi-tenant Support: Supports token passthrough, allowing a single server deployment to serve multiple Notion integrations.
• Broad Client Compatibility: Ready-to-use configurations for Cursor, Claude Desktop, Zed, and GitHub Copilot CLI.