Hosting

Local Python MCP server

Transport

stdio

Endpoint

Local process; no hosted endpoint documented

Authentication

No account or API key is documented for local processing; Whisper model files may download on first use.

Requirements

• Python 3.11 or newer and FFmpeg.
• A local video file; direct YouTube acquisition is listed as future work.

Setup steps

1. Clone the repository, create a Python virtual environment, and run `pip install -e .`.
2. Configure Codex to launch the virtual-environment `video-mcp` executable over stdio.
3. Download a YouTube video separately, then pass its absolute local file path to `process_video`.

Two stored tools retained as matching the repo-level inventory.

• process_video

  Extract scene or interval frames and transcribe a local video into a timeline.

  Parameters: video_path, frame_interval (optional), use_scene_detection (optional), scene_threshold (optional), whisper_model (optional), max_frames (optional)
  01
• list_processed_videos

  List cached processed videos and settings.

  02

• Does not directly fetch YouTube videos according to its README; pair it with a downloader.
• The repository documents token and frame limits primarily for Claude Code, not verified Codex limits.
• Frame sampling and speech transcription do not establish music or sound-effect details.

• maintainer repository readme

  Documents two tools, local setup, scene detection, Whisper, cache, and YouTube as future work.

• creator github avatar

  Local copy decoded as a 300 by 300 JPEG and visually reviewed.

• maintainer repository documentation

  Maintainer repository identifies FFmpeg and Whisper-based video frame and transcription processing; stored two-tool inventory matches the previously checked project interface.

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