Hosting

Local Node.js MCP server

Transport

stdio

Endpoint

Local process; no hosted endpoint documented

Authentication

Local processing needs no API key. OPENAI_API_KEY is an optional cloud transcription fallback; yt-dlp cookies may be configured for restricted platform URLs.

Requirements

• Node.js 18 or newer and FFmpeg.
• yt-dlp for YouTube or other platform URLs.
• Tesseract for OCR and local Whisper for transcription when captions are absent.

Setup steps

1. Clone and build the repository with `npm install && npm run build`, or use the maintainer's release package.
2. Run `video-mcp setup` and `video-mcp doctor` for local dependencies, then `video-mcp install --client codex` to configure Codex.
3. For a YouTube URL, verify yt-dlp is available; use `video_analyze` or the smaller metadata, transcript, and frame tools.

Nine tool rows retained from the implementation source previously recorded.

• video_check_deps

  Check availability of FFmpeg, yt-dlp, and transcription backends.

  01
• video_info

  Probe video duration, resolution, codecs, and audio presence.

  Parameters: source
  02
• video_transcribe

  Get timed speech from captions or a configured Whisper backend.

  Parameters: source
  03
• video_get_frame

  Return one JPEG frame at a requested time.

  Parameters: source, time_sec, inline_image (optional), response_format (optional)
  04
• video_get_frame_burst

  Return multiple frames across a short window.

  Parameters: source, start_sec, end_sec, count (optional), inline_images (optional), response_format (optional)
  05
• video_extract_frames

  Select frames by scene changes, intervals, or both.

  Parameters: source, detail (optional), max_frames (optional), mode (optional), start_sec (optional), end_sec (optional), inline_images (optional), response_format (optional)
  06
• video_analyze

  Combine probe, transcript, key frames, OCR, and timeline.

  Parameters: source
  07
• video_ocr

  Read on-screen text from key frames or one image.

  Parameters: source, max_frames (optional), response_format (optional)
  08
• video_search_transcript

  Find transcript segments matching a query.

  Parameters: source, query
  09

• The maintainer's one-command installer is documented for macOS; Windows support was not independently tested.
• A cloud Whisper fallback is used only if configured; keeping audio local requires a local backend.
• The project also has a shoot-folder sorting CLI, but that CLI command is not an MCP tool.

• maintainer repository readme

  Documents setup, optional yt-dlp, nine tools, local/cloud transcription choice, and Codex configuration.

• maintainer server source

  Registers the nine documented MCP tools and describes their roles.

• project repository icon

  Local SVG decoded and rendered for visual review.

• maintainer repository documentation

  Maintainer repo and source file match the catalog record; stored nine-tool inventory retained as verified from implementation.

Documentation checked 2026-10-28T00:00:00Z. Connection has not been runtime tested.