mcp-video

Video intelligence MCP server for AI-powered video analysis. Extract keyframes, transcripts, and visual content from any video URL — right inside Claude Code or any MCP client.

<p align="center"> </p>

Point it at any video (YouTube, Vimeo, Twitter/X, Instagram, and 1000+ sites) and get:

• Keyframe extraction — scene-change detection picks the most meaningful frames
• Transcript extraction — auto-captions or manual subtitles with timestamps
• Full analysis — frames + transcript together for comprehensive video understanding
• Timestamp snapshots — grab a frame at any specific moment
• Video-to-strategy — extract trading strategies from YouTube/Instagram tutorials and generate runnable Python code (backtrader, zipline, pandas-ta, freqtrade)

Works with Claude's vision capabilities to understand what's actually shown in the video, not just what's said.

Point it at any trading tutorial video and get a complete, runnable strategy:

Convert this trading strategy to code: https://youtube.com/watch?v=...

Extract the RSI strategy from this video and generate backtrader code: https://youtube.com/watch?v=...

It extracts chart patterns, indicators, entry/exit rules from the video frames and transcript, then generates code for your preferred framework.

# Requires Python 3.10+ and ffmpeg
pip install mcp-video

# Or with uv
uv pip install mcp-video

# macOS
brew install ffmpeg yt-dlp

# Ubuntu/Debian
sudo apt install ffmpeg
pip install yt-dlp

# Windows
choco install ffmpeg yt-dlp

Add to your Claude Code MCP config (~/.claude/claude_code_config.json):

{
  "mcpServers": {
    "video": {
      "command": "mcp-video",
      "args": []
    }
  }
}