moshi

GitHub

Moshi is a speech-text foundation model and full-duplex spoken dialogue framework. It uses Mimi, a state-of-the-art streaming neural audio codec.

10,165 stars Python Python 1 file · ~7,851 tokens

moshi MCP Server

MCP server for square/moshi

Ready-to-Use Client Configurations 
{
  "mcpServers": {
    "moshi": {
      "command": "npx",
      "args": [
        "-y",
        "@square/moshi"
      ]
    }
  }
}