What is handy-mcp?

handy-mcp is a tool that lets AI assistants like Claude or Gemini listen to and transcribe audio and video files, all without sending anything to the internet. It works as a Model Context Protocol (MCP) server, which is a standard plug-in format that lets AI tools gain new abilities. In this case, the new ability is speech-to-text: converting recorded speech into readable text, entirely on your own computer. When you connect handy-mcp to your AI tool of choice, you can simply tell the AI to transcribe a meeting recording, summarize what was said in a voice note, or process an entire folder of audio files at once. The AI then uses handy-mcp behind the scenes to do the transcription locally. Because everything happens on your machine, there are no API keys to manage, no usage costs, and your recordings never leave your computer. Under the hood, handy-mcp picks up speech recognition models you have already installed through a companion app called Handy. It supports two families of models: Whisper models, which work through an optimized processing engine called sherpa-onnx, and Parakeet TDT models, which require a custom processing path because the standard engine does not support their architecture. The tool handles multiple file formats including common audio types such as MP3, WAV, M4A, and OGG, and also video files such as MP4, MKV, AVI, and MOV, extracting the audio track automatically. Optional audio enhancement can normalize volume and remove silence before transcription to improve accuracy on noisy recordings. It is written in TypeScript and requires Node.js to run, with an optional dependency on ffmpeg for formats other than WAV.