Yap MCP
Private, on-device speech to text for AI assistants.
Week 15 · Oct 19 – 25, 2026Product
About
Yap MCP is a free, open-source MCP server that lets AI assistants transcribe audio and video files on your own computer. Connect it to Claude, Cursor, VS Code, Windsurf or any other MCP client, and ask the assistant to transcribe a recording. The text comes back without the file ever leaving your machine. Under the hood it uses NVIDIA Parakeet TDT v3, the same speech engine as the Yap dictation app for Mac and Windows. It reads WAV, MP3, M4A, MP4, FLAC, OGG, Opus and WebM files of any length, detects 25 European languages automatically, and can add timestamps to every segment. A read-only search tool also lets the assistant look through the local history of the Yap dictation app. There is no upload, no API key and no account. The speech model is about 670 MB and downloads once, after which everything runs offline. It works on macOS, Windows and Linux with Node.js 20 or newer. Install it with one command: npx -y github:founden-bot/yap-mcp. For Claude Code, run claude mcp add yap -- npx -y github:founden-bot/yap-mcp. Claude Desktop users can also install the one-click bundle from the GitHub releases page. There is also a terminal command, yap-transcribe, for transcribing files straight from the shell. The project is MIT licensed and the full source is on GitHub, so you can read exactly what it does with your files. It is built and maintained by the Yap team at O-mega Enterprise Inc., the makers of Yap voice dictation for Mac and Windows. Good fits: turning meeting recordings into notes inside your coding assistant, transcribing voice memos before summarising them, pulling quotes from interviews, or building a private transcription step into an automation. Because the model runs locally, sensitive recordings stay on your disk. Source code and setup guide: https://github.com/founden-bot/yap-mcp
