mcp-limitless

This MCP server provides advanced analysis of Limitless Pendant conversations, including automatic meeting detection, action item extraction, and comprehensive analytics. It supports a suite of tools for querying, searching, and summarizing lifelogs with consistent pagination to handle large datasets. The server exposes tools such as limitless_detect_meetings, limitless_extract_action_items, limitless_get_raw_transcript, limitless_get_daily_summary, limitless_get_full_transcript, and limitless_get_by_natural_time, among others. To integrate with Claude Desktop or other MCP clients, configure the server under the mcpServers section and supply your Limitless API key. All list-based tools implement cursor-based pagination to avoid token limits, and there are options to fetch complete transcripts when you need full data for downstream processing.

Prerequisites:

• Node.js and npm installed on your system
• Access to the Limitless Pendant API with a valid API key

Installation steps:

1. Install the MCP server globally (example):

npm install -g 199bio-mcp-limitless-server

2. Obtain your Limitless API key from the Limitless service dashboard or via your account portal.
3. Run the MCP server (example):

LIMITLESS_API_KEY=your-key-here limitless-server

4. Alternatively, reference the server via npx in your Claude Desktop configuration:

npx -y 199bio-mcp-limitless-server

5. In Claude Desktop or your MCP client, add the server with the configuration:

{
  "mcpServers": {
    "limitless": {
      "command": "npx",
      "args": ["-y", "199bio-mcp-limitless-server"],
      "env": {
        "LIMITLESS_API_KEY": "your-api-key-here"
      }
    }
  }
}

Tips and notes:

• The MCP server communicates over JSON-RPC via stdio, not HTTP. Do not attempt to connect with a browser or curl.
• Ensure your LIMITLESS_API_KEY is kept secure and never exposed in client-side code.
• Pagination is enabled for all list-based tools; include cursor parameters to fetch subsequent pages.
• If you encounter token-length issues, use the full transcript tool judiciously or request data in smaller chunks.
• When debugging, check that the pendant has recent data and that the API key has the proper permissions for lifelogs, transcripts, and analytics.
• The server supports multiple output modes for the full transcript tool (summary, full, transcript_only) to balance detail with token usage.