YouTube Vision

You can run a YouTube analysis MCP server that leverages the Google Gemini Vision API to describe videos, answer questions, summarize content, and extract key moments. It communicates via standard input/output, making it easy to connect with MCP clients and automate video insights.

How to use

Connect to the YouTube Vision MCP Server from your MCP client and choose the action you want: askAboutYouTubeVideo for descriptions or Q&A, summarizeYouTubeVideo for a concise recap, extractKeyMoments to pull timestamped highlights, or listSupportedModels to see Gemini models that support content generation. Provide the YouTube URL and any optional parameters required by the tool you’re using.

How to install

Prerequisites you need before running the server are: Node.js version 18 or higher and a Google Gemini API key.

{
  "mcpServers": {
    "youtube_vision": {
      "command": "npx",
      "args": [
        "-y",
        "youtube-vision"
      ],
      "env": {
        "GEMINI_API_KEY": "YOUR_GEMINI_API_KEY",
        "GEMINI_MODEL_NAME": "gemini-2.0-flash"
      }
    }
  }
}

Additional installation flow for local build

If you prefer to run from source after cloning, install dependencies, build, and run the compiled code. Use the API key as an environment variable during execution.

# Step 1: Clone the repository
git clone https://github.com/minbang930/Youtube-Vision-MCP.git
cd youtube-vision

# Step 2: Install dependencies
npm install

# Step 3: Build the project
npm run build

# Step 4: Run the compiled server directly (ensure GEMINI_API_KEY is set)
node dist/index.js

Configuration notes

The server uses environment variables to configure the Gemini integration. Set GEMINI_API_KEY to your Google Gemini API key and GEMINI_MODEL_NAME to the desired Gemini model (defaults to gemini-2.0-flash if not specified). You can provide these values in your MCP client's settings under the env section for the youtube_vision server.

Troubleshooting and tips

If you encounter authentication issues, double-check that GEMINI_API_KEY is correct and that the selected GEMINI_MODEL_NAME is not labeled as Experimental or Preview for production use. Ensure your network can reach the Gemini API endpoints and that your API key has the necessary permissions.

Available tools

ask_about_youtube_video

Answers a question about a YouTube video or generates a general description if no question is provided.

summarize_youtube_video

Produces a concise summary of the given YouTube video, with an optional length parameter (short, medium, long).

extract_key_moments

Identifies and outputs key moments from the video with timestamps, up to a specified number of moments.

list_supported_models

Lists Gemini models that support the generateContent method.