Voice MCP

This MCP server provides voice interaction capabilities through multiple transport methods, enabling AI assistants to conduct two-way voice conversations using either local microphone recording or LiveKit room-based communication. Built by mbailey using Python with FastMCP, sounddevice, and OpenAI-compatible APIs, it features automatic transport selection that tries LiveKit first then falls back to local audio, configurable STT/TTS services with support for self-hosted alternatives, debug mode with audio file saving, and privacy-conscious design with clear user consent requirements. The implementation includes tools for asking voice questions with automatic response recording, standalone text-to-speech and speech-to-text functions, audio device detection, and LiveKit room status monitoring, making it valuable for creating voice-enabled AI assistants, accessibility applications requiring audio interaction, and development workflows that need flexible voice processing with both cloud and local deployment options.

Source

Repository: https://github.com/mbailey/voicemode