cartesia-mcp 0.3.0

The official Cartesia MCP server

Cartesia MCP Server

The Cartesia MCP server exposes Cartesia APIs over the Model Context Protocol (MCP) so clients such as Cursor, Claude Desktop, and OpenAI Agents can list voices, run TTS, clone voices, infill audio, and more—without one-off scripts.

Documentation: Cartesia docs — MCP

Requirements

• uv — runs the server via uvx with no global install
• Python 3.13+ (installed automatically by uvx)
• A Cartesia API key

Setup

Add this to your MCP config. You only need your API key.

Cursor — .cursor/mcp.json in your project, or ~/.cursor/mcp.json globally.

Claude Desktop — Settings → Developer → Edit Config (claude_desktop_config.json).

{
  "mcpServers": {
    "cartesia-mcp": {
      "command": "uvx",
      "args": ["cartesia-mcp"],
      "env": {
        "CARTESIA_API_KEY": "<your-api-key>"
      }
    }
  }
}

Restart the client (or refresh MCP in Cursor) and confirm cartesia-mcp is connected.

Try it

Ask your agent things like:

• List all available Cartesia voices
• Convert text to audio with a chosen voice
• Localize an existing voice into another language
• Infill audio between two existing audio segments
• Change an audio file to use a different voice

Tools

Tool	Description
text_to_speech	Convert text to audio with a chosen voice and model
list_voices	List available voices (optionally filter by language or gender)
get_voice	Fetch metadata for a voice by ID
clone_voice	Clone a voice from an audio sample
update_voice	Update a cloned voice's name or description
delete_voice	Delete a cloned voice
infill	Generate audio between two existing audio segments
voice_change	Re-render audio with a different voice
localize_voice	Adapt a voice to another language or dialect

See cartesia_mcp/server.py for parameters and return types.

Testing

Smoke-test all tools (requires CARTESIA_API_KEY):

uv run python scripts/test_all_tools.py

The script creates temporary cloned/localized voices and deletes only those. It does not delete catalog or other existing voices.

Advanced

Output directory

By default, generated audio is written to the server's working directory. To choose a fixed folder, add OUTPUT_DIRECTORY to env:

"env": {
  "CARTESIA_API_KEY": "<your-api-key>",
  "OUTPUT_DIRECTORY": "~/cartesia-output"
}

Local audio files

Tools like infill and voice_change need paths to existing audio files on disk. Pass the full path to each file when prompting your agent.