# video-extract-mcp [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/yanlingLabs/video-extract-mcp  
**GitHub Stars:** 7  
**npm Downloads (last month):** 2043  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/video-extract-mcp

## Description
Extract any video from a URL: transcript plus the keyframes that matter, all locally.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "video-extract-mcp": {
    "command": "npx",
    "args": ["-y","@yanlinglabs/video-extract-mcp@latest"]
  }
}
```

## Documentation

## What video-extract-mcp does

The video-extract-mcp MCP server gives an AI agent a local workflow for working with video links and files. It can inspect a video's metadata without downloading the media, download an entire video or a requested time range, produce a transcript, and return a selected set of keyframes. These operations can be used separately, so downloading is not required before requesting metadata or platform captions.

URL resolution covers named platforms including YouTube, TikTok, Instagram, X, Facebook and Reels, Twitch, Vimeo, Reddit, and WeChat Channels. Direct `.mp4` and `.m3u8` links, generic webpages, and local files are also supported. Unsupported sources return a failure result rather than crashing the process.

## How it works

For metadata-only requests, the server can return details such as the title, creator, duration, chapters, description, and comment count without transferring the video. Downloads can target the complete file or a specified interval. When the source supports ranged fetching, only the requested portion is transferred.

Transcript selection prefers captions supplied by the platform. If captions are unavailable or cannot be retrieved, local speech recognition using Whisper or SenseVoice can transcribe the audio. Results identify the transcript source, and locally generated transcripts include a reason such as missing or failed captions.

Keyframe selection combines scene-boundary detection, blur and quality filtering, on-screen text novelty, subtitle awareness, and image-embedding similarity. The result is intended to represent meaningful visual changes while removing near-duplicates. Uniform sampling and a single frame at an exact timestamp are also available modes.

The video-extract-mcp MCP server writes extracted material to disk and returns a compact response with file paths and summaries. Multiple videos can be submitted together; an error for one item does not invalidate the other results. Longer analyses can run as background tasks, with progress available through a local HTTP endpoint.

## Setup and configuration

The video-extract-mcp MCP server requires Node.js 22.12 or newer, `ffmpeg` with `ffprobe`, `yt-dlp`, Deno, and Tesseract with Chinese language data. The documented tested platform is macOS. Windows and Linux installation instructions are provided separately in the project's installation guide, including PATH and package-version considerations.

The package can be launched through `npx`:

```bash
npx -y @yanlinglabs/video-extract-mcp@latest
```

It can also be installed globally with npm and started through the installed command. Using `@latest` with `npx` allows new releases to be selected on launch; a bare package reference may remain at the first cached version. After installing system dependencies, restart the MCP client so its server process can see them.

Browser cookies can be supplied for media that requires authentication, including age-restricted, members-only, or followers-only content. The provided material does not specify environment variable configuration.

## Tools and capabilities

The server exposes three MCP tools covering the main workflow:

- Resolve and inspect video metadata without downloading the file.
- Download complete videos or selected time ranges, and obtain transcripts from captions or local speech recognition.
- Extract important keyframes, uniformly sampled frames, or a frame at an exact timestamp.

Batch requests can process several videos, and background execution is available for jobs that exceed a client's normal waiting period. Output remains local rather than being sent to a cloud processing service.

## Limitations and notes

Video analysis can take minutes, especially for long files, so MCP client timeouts may need to be increased. Background tasks provide an alternative when a client stops waiting. The required command-line dependencies do not come from npm, and Tesseract's language data may need separate installation.

The project notes that Windows has not yet been run as a tested platform in the supplied documentation. Access to private or rate-limited media depends on usable browser cookies and the source platform. Unsupported URLs return an explicit failure. No claim is made that every website or media format will resolve successfully.

_Full upstream README: https://allmcps.com/mcp/video-extract-mcp/readme_