MLX Whisper MCP Server

Example Use Cases

• 1

  Transcribe audio files on disk into text using a high-quality Whisper model optimized for Apple Silicon.

• 2

  Transcribe uploaded or base64-encoded audio data and receive saved text transcripts alongside the audio.

• 3

  Download YouTube videos and automatically transcribe their audio, with optional retention of the downloaded file.

Description

This MCP server exposes MLX Whisper (mlx-community/whisper-large-v3-turbo) transcription via the Model Context Protocol so clients like Claude Desktop can request transcriptions. It supports local file transcription, base64-encoded audio, and downloading + transcribing YouTube videos, saving .txt transcripts alongside audio files and cleaning up temporary files. Designed for Apple Silicon (M-series), the script auto-manages dependencies via uv and emphasizes easy integration and rich console output for debugging.