ElevenLabs MCP Server

How to use

You connect to the ElevenLabs MCP Server from your MCP client (for example Claude Desktop or similar tools). First, configure your client with the MCP server settings, then issue tasks like generating speech from text, cloning a voice, or transcribing audio. Use the client’s standard workflow to send requests to the MCP server and receive audio or text results. You will typically provide your ElevenLabs API key and select the specific operation you want to perform, such as creating speech in a chosen voice or transforming an existing recording.

How to install

Prerequisites: you need Python available on your system and a way to install the MCP server runtime. You will also need an ElevenLabs API key to access the services.

Install the MCP runtime environment for development and running the server. You will install a Python-based MCP server using the uv package manager as described here.

Install uv if you have not already: run the following command to pull and install uv.

Configuration and quickstart for Claude Desktop

If you plan to use Claude Desktop as your MCP client, follow these steps to configure the ElevenLabs MCP server within Claude Desktop:

1) Open Claude Desktop and go to Settings > Developer > Edit Config > claude_desktop_config.json to include the MCP server configuration shown below. This config runs the MCP server via uvx with your ElevenLabs API key.

{
  "mcpServers": {
    "ElevenLabs": {
      "command": "uvx",
      "args": ["elevenlabs-mcp"],
      "env": {
        "ELEVENLABS_API_KEY": "<insert-your-api-key-here>"
      }
    }
  }
}

Windows note for Claude Desktop

If you are on Windows, you must enable Developer Mode in Claude Desktop to use the MCP server. Open the Help menu (hamburger icon at the top left) and select Enable Developer Mode.

Other MCP clients

For other clients like Cursor and Windsurf, install the Python MCP package and print your configuration to paste into your client.

pip install elevenlabs-mcp
python -m elevenlabs_mcp --api-key={{PUT_YOUR_API_KEY_HERE}} --print

Example usage

Note that ElevenLabs credits are required to use these tools. You can ask for voices, voice variations, or audio transformations, such as turning a recording into a specific character voice or generating a soundscape.

Optional features

You can add the ELEVENLABS_MCP_BASE_PATH environment variable to claude_desktop_config.json to specify the base path MCP server should look for and output files relative to that path.

Contributing and running from source

If you want to contribute or run from source, you can set up a development environment and run tests, then install and run the server through Claude Desktop.

Troubleshooting

Logs when running with Claude Desktop can be found at the following locations: Windows: %APPDATA%\Claude\logs\mcp-server-elevenlabs.log and macOS: ~/Library/Logs/Claude/mcp-server-elevenlabs.log.

Timeouts may occur for long-running operations like voice design or audio isolation when using the MCP inspector in dev mode. These timeouts should not occur when using Claude as the client.

If you see an error such as MCP ElevenLabs: spawn uvx ENOENT, determine the absolute path of uvx on your system and update the configuration. For example, run which uvx to locate it, then set the path in the config, e.g. "command": "/usr/local/bin/uvx".

synthesize_speech

Generate speech audio from text using a selected voice and settings.

clone_voice

Clone or modify a voice profile for reuse across requests.

transcribe_audio

Transcribe speech from an audio input into text.

voice_transform

Transform an existing voice sample to resemble another voice style or character.

audio_scene

Create a soundscape or ambient audio scene to accompany text-to-speech output.