Mcp Listen

Mcp Listen is an open-source toolset enabling AI agents to capture audio, transcribe speech to text, and interact with language models using voice on any MCP-compatible platform. It offers utilities for listing audio devices, real-time or fixed-duration microphone recording, and a 'voice_query' pipeline that transcribes speech locally (with whisper.cpp) and sends queries to a locally running Ollama LLM, returning both transcription and model response. Designed for seamless integration with agents and developer tools, it is privacy-focused—no audio leaves the local device.