spoken-mcp – Multilingual voice interface for Claude Desktop via ElevenLabs

@leszini

Visit Server

a month ago

Spoken-mcp is a voice interface for the Claude Desktop app. It consists of two components: an MCP TTS server that lets Claude speak responses aloud via ElevenLabs, and an STT companion app that transcribes your speech and pastes it into Claude's input field. Supports three input modes: push-to-talk, VAD toggle, and always-on voice detection. No admin rights required. Windows 10/11, Python 3.11+

Overview Tools Comments

Overview

spoken-mcp turns Claude Desktop into a fully voice-enabled assistant. Speak to Claude and hear responses spoken aloud — in any language supported by ElevenLabs. The system consists of two components:

MCP TTS Server: Claude calls the speak tool to read responses aloud via ElevenLabs Text-to-Speech (MP3, 44.1 kHz) STT Companion: A background app that listens to your microphone, transcribes via ElevenLabs Scribe v1, and pastes into Claude's input field

Three input modes: push-to-talk, VAD toggle, and always-on voice detection. System tray icon shows current state (idle/listening/recording). No admin rights required. Requirements: Windows 10/11, Python 3.11+, ElevenLabs API key, Claude Desktop app. Full setup guide and documentation: https://github.com/leszini/spoken-mcp

Try in Playground

Server Config

{
  "mcpServers": {
    "spoken-mcp": {
      "command": "python",
      "args": [
        "C:\\full\\path\\to\\spoken-mcp\\tts_server.py"
      ]
    }
  }
}

Build with ShipAny.