Kokoro Voice

Provides a Kokoro-based MCP server that generates spoken audio from text using the Kokoro model.

python

0

GitHub Stars

python

Language

6 months ago

First Indexed

3 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "noersy-kokoro-voice-mcp-server": {
      "command": "uv",
      "args": [
        "tool",
        "run",
        "mcp-kokoro"
      ]
    }
  }
}

You can run Kokoro-based MCP capabilities locally so AI agents can generate and play spoken audio from text. This MCP server integrates with clients that support the Model Context Protocol, letting you convert text into natural-sounding speech on demand and cache results for faster repeated requests.

How to use

You run this server locally and connect your MCP client to it using the provided MCP endpoints. The server exposes a single primary tool you can invoke: speak. When you ask it to speak, it renders audio from the text you provide and plays it back if your client supports playback. You can customize the voice with a voice name (default is af_heart) and adjust the speaking speed with a speed value (default is 1.0). Generated audio is cached to speed up repeated requests, stored at ~/.cache/mcp_kokoro on your machine.

Typical usage patterns include sending a text prompt to the speak tool to obtain audio for a user interaction, choosing a preferred voice for character consistency, and tuning the speed to suit the application’s tone or accessibility needs. After you set up the MCP connection, your client can request spoken output in real time and benefit from the local audio cache on subsequent requests.

How to install

Prerequisites you need before installing the Kokoro MCP server: Python 3.10 or higher and an audio output device for playback. Some systems may require system dependencies for audio playback (for example, PortAudio). On macOS you can install PortAudio with brew install portaudio. On Linux systems you can install PortAudio libraries with a package manager such as apt.

Install options and commands you can use to set up the server locally:

uv tool install mcp-kokoro

# or, equivalently, using Python's package manager
pip install mcp-kokoro

Additional notes

Caching happens automatically to speed up repeated requests. The cache location is ~/.cache/mcp_kokoro. If you prefer to modify how you access the server, you can adjust your MCP client configuration accordingly.

If you need to run the server in a local, self-contained fashion, you can use either the UV-based approach or the Python package approach shown above. Choose the method that matches your environment and tooling preferences.

Available tools

speak

Generates audio from text and plays it back immediately, with optional voice and speed parameters and automatic caching.