logseq-mcp-server

A read-only MCP server for LogSeq graphs. It gives an AI client 10 tools in four kinds: get (a page, a block, what is open now), find (blocks and pages by text, link, property or period), explore (a page's context, its link network and its timeline) and check (a gate on a [[link]] pass). It never writes to the graph.

This package is the server itself: one native Rust binary, with no Python code and no runtime dependencies. It is the same binary, byte for byte, as the one on the project's GitHub Releases. The wheels are built and attested by the release workflow, attached to the release beside the binaries, and uploaded here from there once the release is public.

Install

Wheels exist for macOS (Apple silicon and Intel) and Linux (x86_64, glibc and musl). There is no Windows wheel yet, so installing on Windows fails with "no matching distribution".

uvx logseq-mcp-server      # run it without installing
uv tool install logseq-mcp-server
pipx install logseq-mcp-server

Set up LogSeq

1. In LogSeq, turn on the HTTP API server (Settings, then API) and generate an auth token.

2. Create ~/.logseq-mcp/config.json:

   {
     "apiUrl": "http://127.0.0.1:12315",
     "authToken": "your-token-here"
   }
   

The server reads its token from that file, so no credentials go in your client's config.

Connect a client

Claude Desktop (claude_desktop_config.json), or any client that starts MCP servers over stdio:

{
  "mcpServers": {
    "logseq": {
      "command": "uvx",
      "args": ["logseq-mcp-server"]
    }
  }
}

Claude Code:

claude mcp add logseq -- uvx logseq-mcp-server

More

The tools, the options (timeoutMs, tips, LOGSEQ_MCP_CONFIG), the Claude Code plugin with skills, and the design notes are in the project README. Report problems on the issue tracker.

Licensed under MIT. The licences of the crates inside the binary are in THIRD-PARTY-NOTICES.txt, installed in the package's .dist-info/licenses directory.