All Products
active open sourcedeveloper tools

ElevenLabs MCP Server

Give AI agents high-fidelity text-to-speech

Links

Install Python package View source

About

Open-source ElevenLabs MCP server for text-to-speech in Claude Code, Claude Desktop, and other MCP clients. Run the Python package with uvx to generate scripted multi-voice audio and list available voices. ElevenLabs account and API usage costs are separate.

Set up ElevenLabs MCP Server

You need

  • uv installed (provides uvx; it downloads elevenlabs-mcp-server 0.1.2 or later and a Python runtime for you)
  • An ElevenLabs API key, set as ELEVENLABS_API_KEY

  1. Check that the server runs

    uvx downloads the published elevenlabs-mcp-server package from PyPI and starts it over stdio. No separate install step is needed.

    ELEVENLABS_API_KEY=your-api-key uvx elevenlabs-mcp-server

  2. Add it to Claude Code

    Register the server once for your user account so every Claude Code project can generate audio and list voices. Replace your-api-key with your own key; keys typed on the command line are saved in your shell history.

    claude mcp add elevenlabs --scope user -e ELEVENLABS_API_KEY=your-api-key -- uvx elevenlabs-mcp-server

  3. Add it to Claude Desktop or another MCP client

    Add this entry to claude_desktop_config.json, or to the MCP settings file of any client that runs stdio servers.

    {
  "mcpServers": {
    "elevenlabs": {
      "command": "uvx",
      "args": ["elevenlabs-mcp-server"],
      "env": {
        "ELEVENLABS_API_KEY": "your-api-key"
      }
    }
  }
}

  4. Seeing an old version? Refresh once

    If uvx keeps starting an older cached copy that fails to start, run this once to download the latest release. Without a key it then exits with "ELEVENLABS_API_KEY environment variable not set", which is expected.

    uvx --refresh elevenlabs-mcp-server

  5. Optional voice settings

    ELEVENLABS_VOICE_ID, ELEVENLABS_MODEL_ID, ELEVENLABS_STABILITY, ELEVENLABS_SIMILARITY_BOOST, and ELEVENLABS_STYLE have defaults; set them in the same env block to change the voice or model. Generated audio is saved to an output folder in the directory where the server starts.

Tech Stack

PythonMCP ProtocolElevenLabs API

Available On

pypi

Pricing

Open Source (Free)