mcp-elevenreader

MCP server for ElevenReader — ElevenLabs text-to-speech reader.

Setup

1. Get refresh token

Open https://elevenreader.io, log in, then run in browser console (F12 → Console):

JSON.parse(localStorage.getItem(Object.keys(localStorage).find(k => k.startsWith('firebase:authUser:')))).stsTokenManager.refreshToken

The refresh token is long-lived (months). Access tokens are refreshed automatically.

2. Connect to Claude Code

claude mcp add elevenreader -e ELEVEN_REFRESH_TOKEN=your-token -- uvx mcp-elevenreader

To make it available in all projects, add to ~/.claude/mcp.json:

{
  "mcpServers": {
    "elevenreader": {
      "command": "uvx",
      "args": ["mcp-elevenreader"],
      "env": {
        "ELEVEN_REFRESH_TOKEN": "your-token"
      }
    }
  }
}

Other clients (Claude Desktop / Kiro / Cursor)

{
  "mcpServers": {
    "elevenreader": {
      "command": "uvx",
      "args": ["mcp-elevenreader"],
      "env": {
        "ELEVEN_REFRESH_TOKEN": "your-token"
      }
    }
  }
}

Tools

Architecture

• Auth: Firebase refresh token → short-lived access token (auto-refreshed, thread-safe)
• Data source: /v1/reader/collections/books endpoint (full history, 345+ books)
• Caching: 60s TTL on book list, invalidated on mutations (add/delete)
• Upload queue: Background thread with retry (3 attempts), rate limiting, pause for priority uploads
• Thread safety: Locks on token cache and reads cache

Development

git clone https://github.com/MIt9/mcp-elevenreader
cd mcp-elevenreader
uv sync
mcp dev src/mcp_elevenreader/server.py

Requirements

• Python ≥ 3.10
• Dependencies: httpx, mcp

License

MIT