ElevenLabs MCP Server

How to use

You connect to the ElevenLabs MCP Server from your MCP client by configuring the client to run the server locally. The server provides a set of tools you can invoke from the client to generate speech, clone voices, and transcribe audio. In practice, you install the MCP runtime, supply your API key, and then point your client’s configuration at the local MCP server so you can start issuing tool requests from the client’s interface.

How to install

Prerequisites: you need a Python-friendly environment and a runtime that supports MCP. You will also need an ElevenLabs API key to access the services.

Step 1: Install the MCP runner for your environment. If you use the UV package manager you install it with the provided script, then use UV to install Python tooling for the MCP server.

# Install UV (Python package manager)
curl -LsSf https://astral.sh/uv/install.sh | sh

# (Optional) Create and activate a virtual environment if you prefer isolated installs
uv venv
source .venv/bin/activate

# Install the ElevenLabs MCP server package (Python) in editable mode if you’re developing
uv pip install -e ".[dev]"

Step 2: Obtain your API key from ElevenLabs and configure the MCP server in your client. The standard approach is to run the MCP server as a local stdio process and provide the API key through environment variables.

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

Step 3: Start the MCP server via your MCP client configuration. For Claude Desktop and other supported clients, place the above configuration in the client’s MCP settings and ensure the API key is active. If you are on Windows, enable Developer Mode in Claude Desktop to use the MCP server.

Additional features and notes

Optional features let you control how generated files are returned or stored. You can adjust how outputs are delivered using environment variables in your client configuration. The available options include returning files directly, returning them as MCP resources, or returning both.

"env": {
  "ELEVENLABS_API_KEY": "your-api-key",
  "ELEVENLABS_MCP_OUTPUT_MODE": "files"  // or "resources" or "both"
}

Data residency can be set to choose the geographic region for data processing. This is an enterprise feature and requires appropriate permissions.

Troubleshooting and tips

If you encounter timeouts during long-running API operations, this is often caused by client-side timeouts rather than server processing. Use a client like Claude to avoid premature timeouts during extended tasks.

If you see an error about spawning uvx, locate the absolute path to the executable and update your configuration to reference that path, for example: "command": "/usr/local/bin/uvx".

Tools and capabilities overview

The ElevenLabs MCP Server exposes a set of tools that enable core audio tasks: generate speech from text, clone voices, transcribe audio, and apply audio processing to create soundscapes or modify voice characteristics.

Security and access

Keep your API key secure. Do not commit keys into configuration files that may be shared or stored in public repositories. Use environment variables or secret management features provided by your deployment environment.

Notes on contributing and running locally

If you want to develop or run from source, clone the project, set up a virtual environment, install dependencies, and run tests to verify the setup. You can then install and debug the server within your MCP-enabled client environment.

synthesize_speech

Generate speech from text using ElevenLabs TTS with MCP integration.

clone_voice

Clone an existing voice profile for custom speech rendering.

transcribe_audio

Transcribe uploaded audio into text with speaker labeling if available.

audio_processing

Apply audio processing tasks such as soundscapes or effects to input audio.