MCPTube

Install the mcptube package via pip and ensure ffmpeg is available on your system for scene-change frame detection.

pip install mcptube
# macOS: brew install ffmpeg
# Ubuntu/Debian: sudo apt install ffmpeg
# Windows: download from ffmpeg.org and add to PATH

MCPTube auto-detects which LLM provider to use based on which API key is set. Set at least one of the following environment variables. In MCP client config mode, set these in the env block.

export OPENAI_API_KEY=your-openai-api-key
# or: export ANTHROPIC_API_KEY=your-anthropic-api-key
# or: export GOOGLE_API_KEY=your-google-api-key

Register mcptube as an stdio MCP server in your client configuration. Pass your API key via the env block.

{
  "mcpServers": {
    "mcptube": {
      "command": "python",
      "args": ["-m", "mcptube"],
      "env": {
        "OPENAI_API_KEY": "your-openai-api-key"
      }
    }
  }
}

Ask your AI client to add a YouTube video to the knowledge base. The server will extract the transcript, detect scene changes with ffmpeg, run vision analysis on key frames, and write wiki pages to disk.

After ingesting videos, ask questions. The agent uses FTS5 search to find relevant wiki pages, then reasons over them with full structural context — citations point back to specific videos and timestamps.