ai-bookmarks-mcp

Model Context Protocol server that exposes your Chrome bookmarks to AI agents — Claude Desktop, Antigravity, Cursor, Cline, or any other MCP client. Ask your agent to find bookmarks by topic, list your folder taxonomy, or look up a specific saved page.

Companion to AI Bookmark Manager, a Chrome extension that tags, summarizes, and organizes bookmarks with AI — but this server works standalone: it reads Chrome's native Bookmarks file directly, no extension required.

Tools

• search_bookmarks — keyword search across title/URL/summary/tags, scored by how many query words match. Since this is meant to be called by an LLM agent rather than typed directly, the agent can rephrase or retry with synonyms itself when a query comes up empty.
• get_all_bookmarks — every bookmark at once (id, title, url, tags, summary, folderId), no search filtering, optionally scoped to one folder.
• list_folders — the full folder hierarchy and each folder's AI prompt context.
• get_bookmark — full details (tags, summary, dates) for one bookmark by ID or URL.

Setup

No installation needed — run it with npx:

{
  "mcpServers": {
    "ai-bookmarks": {
      "command": "npx",
      "args": ["-y", "ai-bookmarks-mcp"]
    }
  }
}

Paste that into:

• Claude Desktop: %APPDATA%\Claude\claude_desktop_config.json
• Antigravity: your user or project mcp_config.json
• Cursor / Cline: your mcp.json

By default it auto-detects your Chrome (or Edge/Brave) Bookmarks file. To point it at a specific file instead, set BOOKMARKS_DATA_PATH:

{
  "mcpServers": {
    "ai-bookmarks": {
      "command": "npx",
      "args": ["-y", "ai-bookmarks-mcp"],
      "env": { "BOOKMARKS_DATA_PATH": "C:\\path\\to\\Bookmarks" }
    }
  }
}

License

MIT