Mcp Listen is an MCP server that adds listening capability and speech-to-text tools to AI assistants and agents. It exposes tools to list microphone devices, capture audio from a mic, and provide end-to-end offline voice query pipelines: capturing speech, transcribing it locally (via whisper.cpp), and querying a local LLM (Ollama). The server is designed for developers and users building custom voice-enabled AI agents or integrating voice inputs into MCP-compatible clients (Claude, Cursor, Windsurf, Gemini, VS Code, etc.), with zero reliance on cloud services or external APIs for audio or transcription.
Visit Mcp Listen's official website for product details and getting started.