Cartesia MCP Server

Cartesia MCP Server

What is Cartesia MCP Server about?

Provides a local MCP (Model Context Protocol) server that bridges applications such as Claude Desktop, Cursor, and OpenAI agents with Cartesia's speech AI platform. It supports listing voices, converting text to audio, localizing speech to different languages, infilling missing audio segments, and re‑voicing existing files.

How to use Cartesia MCP Server?

1. Create a Cartesia account and obtain an API key from the Cartesia playground.
2. Install the package:

   pip install cartesia-mcp
   which cartesia-mcp   # get the absolute path to the executable
   

3. Configure the client (e.g., Claude Desktop or Cursor) by referencing the executable path and supplying CARTESIA_API_KEY (and optionally OUTPUT_DIRECTORY).
4. Run the server and issue MCP commands from the integrated client to perform audio operations.

Key Features

• Voice listing – Retrieve all available Cartesia voices.
• Text‑to‑audio – Generate speech from any text using a chosen voice.
• Localization – Translate and synthesize speech in a different language while preserving the speaker's style.
• Audio infill – Seamlessly generate audio between two existing clips.
• Voice swapping – Change the voice of an existing audio file to another Cartesia voice.
• Simple integration – Ready‑to‑use config snippets for Claude Desktop and Cursor.

Use Cases

• Developers building AI assistants that need on‑demand speech synthesis.
• Content creators generating multilingual voice‑overs.
• Applications requiring dynamic audio generation such as interactive games or virtual agents.
• Automated pipelines that need to stitch or replace audio segments programmatically.

FAQ

Q: Do I need a paid Cartesia account? A: No, a free tier provides 20,000 credits per month, sufficient for most development and testing.

Q: Which languages are supported for localization? A: All languages offered by Cartesia’s API can be used; refer to the Cartesia documentation for the full list.

Q: Can I run the server without installing Python packages? A: The server is distributed as a Python package; installation via pip is required.

Q: How do I specify the output folder for generated audio? A: Set the OUTPUT_DIRECTORY environment variable in the client configuration; if omitted, files are saved in the current working directory.

Cartesia MCP Server

The Cartesia MCP server provides a way for clients such as Cursor, Claude Desktop, and OpenAI agents to interact with Cartesia's API. Users can localize speech, convert text to audio, infill voice clips etc.

Cartesia Setup

Ensure that you have created an account on Cartesia, there is a free tier with 20,000 credits per month. Once in the Cartesia playground, create an API key under API Keys --> New.

Installation

pip install cartesia-mcp
which cartesia-mcp # absolute path to executable

Claude Desktop Integration

Add the following to claude_desktop_config.json which can be found through Settings --> Developer --> Edit Config.

{
  "mcpServers": {
    "cartesia-mcp": {
      "command": "<absolute-path-to-executable>",
      "env": {
        "CARTESIA_API_KEY": "<insert-your-api-key-here>",
        "OUTPUT_DIRECTORY": // directory to store generated files (optional)
      }
    }
  }
}

Try asking Claude to

• List all available Cartesia voices
• To convert a text phrase into audio using a particular voice
• To localize an existing voice into a different language
• To infill audio between two existing audio segments (specify absolute paths to audio files)
• To change an audio file to use a different voice

Cursor Integration

Create either a .cursor/mcp.json in your project or a global ~/.cursor/mcp.json. The same config as for Claude can be used.