Skip to main content
MCP server CX Codex

Video Extract — Claude Code MCP server: what it does, how to install

Video Extract — an MCP server that downloads videos from YouTube, TikTok, WeChat and hundreds of other platforms, extracts transcripts and keyframes. Everything runs locally, no API keys.

MIT ★ 12 updated 2026-10-21

Install into Claude Code · Platform CLI All methods and platforms ↓
claude mcp add --env VIDEO_EXTRACT_COOKIES_FROM_BROWSER=<VIDEO_EXTRACT_COOKIES_FROM_BROWSER> --env VIDEO_EXTRACT_MODELS_DIR=<VIDEO_EXTRACT_MODELS_DIR> --env VIDEO_EXTRACT_MAX_CONCURRENCY=<VIDEO_EXTRACT_MAX_CONCURRENCY> --transport stdio video-extract -- npx -y @yanlinglabs/video-extract-mcp
Open on GitHub

What Video Extract does

The server takes a video link — YouTube, TikTok, Instagram, X, Facebook, Twitch, Vimeo, Reddit, WeChat Channels, a direct .mp4 or .m3u8 — and downloads the entire file or just the section you specify by timecode. It then extracts a transcript: first attempts to fetch published captions from the platform, and if none exist, performs local speech recognition via Whisper or SenseVoice. Keyframes are selected by scene boundaries, sharpness, on-screen text appearance and visual similarity — instead of thousands of near-identical stills, the agent receives a few dozen meaningful frames.

For private or age-restricted videos, you can provide browser cookies — the server will pass access checks and download content unavailable anonymously. Long tasks run in the background and report progress over HTTP, even if the client stops waiting.

Requires ffmpeg, ffprobe, yt-dlp, Deno, tesseract with Chinese language data, and Node ≥ 22.12 in PATH. Without them, the server will not start.

What it helps with

  • download videos from YouTube, TikTok, Instagram, WeChat Channels and hundreds of other platforms
  • get a video transcript — published captions or local speech recognition
  • extract keyframes by scene changes and text appearance, without duplicates
  • download private or restricted videos using browser cookies
  • run long video analysis in the background with progress reports

Example prompts for the agent

  • Download the video from this link and give me the transcript
  • Extract keyframes from the video in lecture.mp4
  • Get video information — title, creator, duration and chapter list — without downloading

How to install Video Extract

Platform CLI

claude mcp add --env VIDEO_EXTRACT_COOKIES_FROM_BROWSER=<VIDEO_EXTRACT_COOKIES_FROM_BROWSER> --env VIDEO_EXTRACT_MODELS_DIR=<VIDEO_EXTRACT_MODELS_DIR> --env VIDEO_EXTRACT_MAX_CONCURRENCY=<VIDEO_EXTRACT_MAX_CONCURRENCY> --transport stdio video-extract -- npx -y @yanlinglabs/video-extract-mcp

Replace <…> with your own values

Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more

Questions about Video Extract

How to install Video Extract in Claude Code?

Run claude mcp add video-extract-mcp -- npx -y @yanlinglabs/video-extract-mcp. Before that, install ffmpeg, yt-dlp, tesseract with language data, and Node ≥ 22.12 via Homebrew (macOS) or winget (Windows), then restart the agent.

What does Video Extract do?

Downloads videos from YouTube, TikTok, Instagram, WeChat and other platforms, extracts transcripts — published captions or local recognition — and selects keyframes by scene changes and text appearance. Everything happens on your machine, no cloud APIs.

What programs does Video Extract need?

ffmpeg with ffprobe, yt-dlp, Deno, tesseract with Chinese language data, and Node ≥ 22.12. They must be in PATH — the server checks for them at startup and will not run without them.

Similar MCP servers

All MCP servers for AI and ML →

Project sponsors

Companies that keep this analytics open