A speech-to-text MCP server for Claude, Codex and Gemini CLI
Short answer: run uvx babelscribe mcp as a local MCP server. Your AI app can then find a video on your computer, transcribe it on your own GPU and write .srt subtitles next to it — nothing is uploaded.
Install
# Claude Code
claude mcp add babelscribe -- uvx babelscribe mcp
# OpenAI Codex CLI
codex mcp add babelscribe -- uvx babelscribe mcp
# Claude Desktop: download babelscribe.mcpb from the GitHub releases and double-click it
For Gemini CLI, Cursor, Google Antigravity and VS Code, add {"command": "uvx", "args": ["babelscribe", "mcp"]} to the app's MCP config; the README has the exact file for each app.
Tools
| Tool | What it does |
|---|---|
transcribe | Audio or video file to subtitles (SRT, VTT) and text, 99 languages |
find_media | Finds the newest audio/video in Downloads, Videos, Desktop |
list_devices | Shows the GPUs available |
list_languages_and_models | Lists languages, models and fine-tunes |
Try it
Ask in plain words: “make Thai subtitles for the newest video in my Downloads”, then “proofread the names” or “translate to English”. Set the tool timeout to 3600 seconds for long videos (Codex defaults to 60 s).
FAQ
Is there an MCP server for speech to text?
Yes. babelscribe is a local MCP server (uvx babelscribe mcp) that transcribes audio and video with Whisper on your own GPU.
Can Claude transcribe a video on my computer?
Yes. Add babelscribe as an MCP server in Claude Desktop (one-click .mcpb) or Claude Code, then ask for subtitles in chat.
Does it upload my files?
No. Transcription runs locally; there is no telemetry and no cloud call.
Can I use it from ChatGPT or Grok on the web?
No. Web chat apps can only reach servers on the internet, not your computer. Use Claude Desktop, Claude Code, Codex CLI, Gemini CLI or Cursor.
babelscribe is free and open source (MIT) · Privacy