The API works asynchronously — submit a video, then collect the transcript, translation, summary or subtitles as they become ready. The practical consequence is spelled out rather than hidden: these calls can run longer than a debugging tool's default timeout, which is exactly the kind of detail that costs an afternoon when it is not documented. Search within a video is the underrated tool here — it finds the moment instead of handing back the whole transcript.
An MCP server over the YouTube Translate API. It goes past the transcript: the same video can be translated into another language, rendered as subtitle files, summarised, or searched for a specific moment.
- Transcripts of YouTube videos, pulled by URL or by the video ID directly
- Translation of a transcript into another language
- Subtitle generation in SRT or VTT format
- Summaries of a video's content
- Search within a video's content, for finding where something is said rather than reading the whole transcript
An API key for the YouTube Translate API, supplied as `YOUTUBE_TRANSLATE_API_KEY` — required. Python 3.12 or higher; a Docker image is provided. Two transports: stdio by default, or `--transport sse --port 8000`. Note that video processing calls can outlast the MCP Inspector's built-in timeout, so debugging is easier against a real client and its logs. The pyproject names the package `youtube-translate-mcp` at 0.1.0, which PyPI does not have.
One command plus a key — npx -y @smithery/cli install @brianshin22/youtube-translate-mcp --client claude, then supply credentials