YouTube Gemini MCP Server

A Model Context Protocol (MCP) server that provides conversational YouTube video analysis capabilities using Gemini 2.5 Pro. Analyze videos through multi-turn conversations with advanced visual understanding.

🎯 What Problems Does This Solve?

Traditional Video Analysis Pain Points

❌ Transcript-only limitations - Missing visual context and scene understanding
❌ No conversation memory - Each analysis starts from scratch
❌ Manual workflows - Complex download/upload processes

Our Solution

✅ Full video analysis - Visual, audio, and contextual understanding
✅ Session memory - Builds on previous analysis automatically
✅ Integrated workflows - YouTube URL to analysis in one step

🚀 Key Capabilities

🔄 Session-Based Video Conversations

Create persistent sessions for multi-turn video analysis:

# Create session
create_video_session(
    description="Analyze machine learning lecture",
    video_source="https://youtube.com/watch?v=abc123",
    source_type="youtube_url"
)

# Continue conversation
analyze_video_in_session(session_id="uuid", prompt="What are the key concepts?")
analyze_video_in_session(session_id="uuid", prompt="Explain the neural networks part")

📹 Direct YouTube Processing

No downloads required - process YouTube videos directly:

analyze_youtube_video(
    youtube_url="https://youtube.com/watch?v=abc123",
    prompt="Summarize this video's main points"
)

💾 Local Video Support

Upload and analyze local video files (48-hour retention):

analyze_local_video(
    video_path="/path/to/video.mp4",
    prompt="What happens in this video?"
)

🛠️ Installation

Prerequisites

• Python 3.10+
• Google AI API key
• Poetry (recommended) or pip

Setup

# Clone repository
git clone https://github.com/aigentive/youtube-gemini-mcp
cd youtube-gemini-mcp

# Install with Poetry
poetry install

# Or with pip
pip install -e .

# Set environment variable
export GOOGLE_API_KEY="your_google_api_key_here"

🔧 Claude Desktop Integration

Quick Setup

1. Get Google AI API Key: Visit Google AI Studio to get your free API key
2. Add to Claude Desktop: Copy the configuration below to your Claude Desktop settings

Production Configuration

Add to your Claude Desktop MCP configuration (claude_desktop_config.json):

{
  "mcpServers": {
    "youtube-gemini-mcp": {
      "command": "youtube-gemini-mcp",
      "env": {
        "GOOGLE_API_KEY": "your_google_api_key_here"
      }
    }
  }
}

Development Configuration

For development work, use the Poetry configuration:

{
  "mcpServers": {
    "youtube-gemini-mcp-dev": {
      "command": "poetry",
      "args": ["run", "python", "-m", "youtube_gemini_mcp.server"],
      "cwd": "/absolute/path/to/youtube-gemini-mcp",
      "env": {
        "GOOGLE_API_KEY": "your_google_api_key_here",
        "LOG_LEVEL": "DEBUG"
      }
    }
  }
}

Configuration Templates

We provide ready-to-use configuration files:

• Development: mcp-config.poetry.json - For local development
• Production: mcp-config.private.json.example - Copy and customize

📚 Available Tools

Tool	Description
create_video_session	Create new conversational analysis session
analyze_video_in_session	Analyze video within session context
analyze_youtube_video	Single-shot YouTube video analysis
analyze_local_video	Single-shot local video analysis
get_session_status	Get session information and history
list_active_sessions	List all active sessions
close_session	Close session and cleanup resources
validate_youtube_url	Validate and normalize YouTube URLs
get_usage_guide	Comprehensive documentation
get_server_stats	Server health and statistics

🎯 Use Cases

Educational Content Analysis

Analyze lectures, tutorials, and educational videos with follow-up questions.

Content Research

Research documentaries, interviews, and informational content systematically.

Video Summarization

Extract key insights and create summaries from long-form content.

Training Material Development

Analyze existing training videos to extract learning objectives and key points.

⚡ Quick Examples

Analyze a YouTube Video

# Direct analysis - perfect for quick insights
analyze_youtube_video(
    youtube_url="https://youtube.com/watch?v=dQw4w9WgXcQ",
    prompt="What is this video about? Provide a comprehensive summary."
)

Session-Based Analysis

# Create persistent session for multi-turn conversation
session = create_video_session(
    description="Learning about machine learning fundamentals",
    video_source="https://youtube.com/watch?v=abc123",
    session_name="ML Fundamentals Study"
)

# Build understanding through conversation
analyze_video_in_session(
    session_id=session["session_id"], 
    prompt="What are the main topics covered in this lecture?"
)

analyze_video_in_session(
    session_id=session["session_id"], 
    prompt="Focus on the neural networks section - how are they explained?"
)

analyze_video_in_session(
    session_id=session["session_id"], 
    prompt="What practical examples or demonstrations are shown?"
)

analyze_video_in_session(
    session_id=session["session_id"], 
    prompt="Based on our discussion, what are the key takeaways for beginners?"
)

Local Video Analysis

# Upload and analyze private content
analyze_local_video(
    video_path="/path/to/your/presentation.mp4",
    prompt="Extract the key business metrics and recommendations from this quarterly review"
)

Advanced: Timestamp-Specific Analysis

# Focus on specific video segments
analyze_video_in_session(
    session_id="your-session-id",
    prompt="Analyze the demonstration shown in this segment",
    timestamp_range="5:30-8:45"
)

🔐 Configuration

Environment Variables

Required:

• GOOGLE_API_KEY - Get your free API key from Google AI Studio

Optional Configuration:

MCP_MAX_SESSIONS=50               # Maximum concurrent sessions
MCP_SESSION_TIMEOUT=7200          # Session timeout in seconds (2 hours)
GEMINI_MODEL_DEFAULT="gemini-2.5-pro-preview-05-06"  # Default Gemini model
MAX_VIDEO_DURATION=7200           # Max video length in seconds
AUTO_CLEANUP_FILES=true           # Auto-cleanup uploaded files
LOG_LEVEL=INFO                    # Logging level (DEBUG, INFO, WARNING, ERROR)

System Requirements

• Python: 3.10+ (3.11+ recommended)
• Memory: 4GB RAM minimum for video processing
• Storage: 500MB for session data and temporary files
• Network: High-speed internet for video processing

Limitations & Constraints

YouTube Videos (Unlimited Sessions)

• ✅ No file size limits - Direct URL processing
• ✅ Unlimited session duration - No 48-hour restriction
• ⚠️ 2-hour video limit - Recommended maximum for optimal performance
• ⚠️ Public videos only - Private/unlisted may not be accessible

Local Videos (48-Hour Sessions)

• ⚠️ 2GB file size limit - Google Files API restriction
• ⚠️ 48-hour auto-deletion - Files automatically deleted after 48 hours
• ⚠️ 20GB project limit - Total storage quota per Google project
• ❌ No retention extension - Cannot extend file lifespan beyond 48 hours

Performance Guidelines

• Concurrent sessions: Default maximum 50 active sessions
• Session timeout: 2 hours of inactivity before cleanup
• Memory usage: ~100MB per active session
• Response time: 10-30 seconds for typical analysis requests

🧪 Testing

# Run tests
poetry run pytest

# With coverage
poetry run pytest --cov=youtube_gemini_mcp

# Type checking
poetry run mypy src/

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🔗 Related Projects

• MCP Specification
• Claude Desktop
• Google Gemini API

📚 Documentation

• LLM.md - Comprehensive usage guide for AI systems
• DEVELOPMENT.md - Development setup and architecture
• CONTRIBUTING.md - How to contribute to the project
• PRD.md - Complete product requirements document

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

Quick Start for Contributors

1. Fork the repository
2. Set up development environment: poetry install
3. Run tests: poetry run pytest
4. Make your changes and add tests
5. Submit a pull request

📞 Support & Community

• 🐛 Bug Reports: GitHub Issues
• 💡 Feature Requests: GitHub Issues
• 💬 Discussions: GitHub Discussions
• 📖 Documentation: Check our comprehensive docs above

Getting Help

1. Check Documentation: Start with LLM.md for usage questions
2. Search Issues: Look for existing solutions in GitHub Issues
3. Create Issue: Provide clear details and reproduction steps
4. Community: Join discussions for general questions and ideas