Gives Claude microphone access with three tools: list audio devices, capture raw WAV files to disk, and run a full offline voice pipeline. The voice_query tool chains local whisper.cpp transcription with Ollama, so you can speak a question and get an LLM response without anything leaving your machine. Built on decibri for cross-platform audio capture with no ffmpeg dependencies. You specify recording duration up front since there's no VAD stop detection. Useful if you want voice input in Claude Desktop or need to prototype voice workflows that stay local.
claude mcp add --transport stdio analyticsinmotion-mcp-listen uvx mcp-listen