Google AI MCP Server

A comprehensive Model Context Protocol (MCP) server that integrates Google's cutting-edge AI services (VEO 3, Imagen 4, Gemini, and Lyria 2) with Anthropic's Claude Desktop application.

🌟 Features

• 🎬 VEO 3 Video Generation: Create high-quality videos with native audio (5-8 seconds)
• 🎨 Imagen 4 Image Generation: Generate photorealistic images with style control
• 💬 Gemini Text Generation: Access Google's latest LLMs (1.5 Pro, Flash, 2.0 Flash)
• 🎵 Lyria 2 Music Generation: Create studio-quality music (up to 60 seconds)
• ⚡ Batch Processing: Process multiple generations in parallel
• 💰 Cost Estimation: Know costs before making API calls
• 🛡️ Quota Management: Smart rate limiting and usage tracking
• 💾 Smart Caching: Reduce costs with intelligent response caching

📋 Prerequisites

• Node.js 18 or higher
• Google Cloud Project with billing enabled
• Vertex AI API enabled
• Service account with appropriate permissions
• VEO 3 allowlist access (request through Google Form)

🚀 Quick Start

Option 1: Install via Smithery (Coming Soon)

Once approved on Smithery:

npx @smithery/cli install --client claude @stevekaplanai/google-ai-mcp

Option 2: Install via npm

# Install globally
npm install -g @stevekaplanai/google-ai-mcp

# Or use directly with npx
npx @stevekaplanai/google-ai-mcp

Option 3: Build from source

# Clone and build from source
git clone https://github.com/Stevekaplanai/google-ai-mcp-server.git
cd google-ai-mcp-server
npm install
npm run build

2. Google Cloud Setup

For testing with Steve's project (starry-center-464218-r3):

# Set your project
gcloud config set project starry-center-464218-r3

# Enable required APIs
gcloud services enable aiplatform.googleapis.com

# Create service account (if not already done)
gcloud iam service-accounts create mcp-vertex-ai \
  --display-name="MCP Vertex AI Service Account"

# Grant necessary permissions
gcloud projects add-iam-policy-binding starry-center-464218-r3 \
  --member="serviceAccount:[email protected]" \
  --role="roles/aiplatform.user"

# Generate key
gcloud iam service-accounts keys create credentials.json \
  --iam-account=mcp-vertex-ai@starry-center-464218-r3.iam.gserviceaccount.com

For general setup:

# Create service account
gcloud iam service-accounts create mcp-vertex-ai \
  --display-name="MCP Vertex AI Service Account"

# Grant necessary permissions
gcloud projects add-iam-policy-binding YOUR_PROJECT_ID \
  --member="serviceAccount:mcp-vertex-ai@YOUR_PROJECT_ID.iam.gserviceaccount.com" \
  --role="roles/aiplatform.user"

# Download key
gcloud iam service-accounts keys create key.json \
  --iam-account=mcp-vertex-ai@YOUR_PROJECT_ID.iam.gserviceaccount.com

3. Configuration

Set environment variables:

export GOOGLE_CLOUD_PROJECT="your-project-id"
export GOOGLE_CLOUD_LOCATION="us-central1"
export GOOGLE_APPLICATION_CREDENTIALS_JSON='{"type":"service_account",...}'

Or create a .env file:

GOOGLE_CLOUD_PROJECT=your-project-id
GOOGLE_CLOUD_LOCATION=us-central1
GOOGLE_APPLICATION_CREDENTIALS_JSON='{"type":"service_account",...}'

4. Claude Desktop Integration

Add to your Claude Desktop configuration:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "google-ai": {
      "command": "npx",
      "args": ["@stevekaplanai/google-ai-mcp"],
      "env": {
        "GOOGLE_CLOUD_PROJECT": "your-project-id",
        "GOOGLE_CLOUD_LOCATION": "us-central1",
        "GOOGLE_APPLICATION_CREDENTIALS_JSON": "{...}"
      }
    }
  }
}

For testing with mock mode:

{
  "mcpServers": {
    "google-ai": {
      "command": "npx",
      "args": ["@stevekaplanai/google-ai-mcp"],
      "env": {
        "USE_MOCK": "true"
      }
    }
  }
}

🛠️ Available Tools

veo_generate_video

Generate videos using Google VEO 3.

Parameters:

• prompt (string): Text description of the video
• duration (number): Length in seconds (5-8)
• aspectRatio (string): Video aspect ratio (16:9, 9:16, 1:1)
• imageBase64 (string, optional): Base64 image for image-to-video
• sampleCount (number): Number of videos to generate (1-4)
• negativePrompt (string, optional): What to avoid in generation
• outputStorageUri (string, optional): GCS bucket for output

Example:

{
  "prompt": "A serene sunrise over mountains with mist",
  "duration": 8,
  "aspectRatio": "16:9"
}```

### `imagen_generate_image`
Generate images using Google Imagen 4.

**Parameters:**
- `prompt` (string): Text description of the image
- `sampleCount` (number): Number of images (1-8)
- `aspectRatio` (string): Image aspect ratio
- `negativePrompt` (string, optional): What to avoid
- `language` (string): Prompt language (default: 'en')

### `gemini_generate_text`
Generate text using Gemini models.

**Parameters:**
- `prompt` (string): Input text prompt
- `model` (string): Model to use (gemini-1.5-pro, gemini-1.5-flash, gemini-2.0-flash-exp)
- `temperature` (number): Controls randomness (0-2)
- `maxTokens` (number): Maximum tokens to generate
- `topP` (number): Nucleus sampling parameter
- `topK` (number): Top-K sampling parameter

### `lyria_generate_music`
Generate music using Lyria 2.

**Parameters:**
- `prompt` (string): Music description- `duration` (number): Length in seconds (5-60)
- `genre` (string, optional): Music genre
- `mood` (string, optional): Music mood
- `instruments` (array, optional): Instruments to include

### `check_operation_status`
Check the status of long-running operations.

**Parameters:**
- `operationName` (string): Operation ID from generation request
- `modelType` (string): Type of model (veo, imagen, lyria)

## 📚 Resources

The server provides access to these resources:

- `google-ai://models` - Information about available models and capabilities
- `google-ai://pricing` - Current pricing information for each service

## 💡 Usage Examples

### In Claude Desktop

Use veo_generate_video to create a 5-second video of "A futuristic city at sunset with flying cars"

Generate 4 product photos of "A modern minimalist watch on white background" using imagen_generate_image

Write a blog post about sustainable energy using gemini_generate_text with model gemini-1.5-pro

Create 30 seconds of upbeat electronic music using lyria_generate_music

## 🔧 Advanced Features

### Batch Processing
Process multiple generations in parallel:
```javascript
// Coming soon: Batch API support

Cost Estimation

Get cost estimates before generation:

// Integrated cost calculation based on current pricing

Quota Management

Built-in rate limiting to prevent API overages:

• VEO: 5 requests/minute
• Imagen: 20 requests/minute
• Gemini: 60 requests/minute
• Lyria: 3 requests/minute

🐛 Troubleshooting

"Permission denied" error

Ensure your service account has the roles/aiplatform.user role:

gcloud projects add-iam-policy-binding YOUR_PROJECT_ID \
  --member="serviceAccount:YOUR_SERVICE_ACCOUNT@YOUR_PROJECT_ID.iam.gserviceaccount.com" \
  --role="roles/aiplatform.user"

"VEO 3 not available" error

VEO 3 requires allowlist access. Request access through the Google Form.

"Invalid location" error

Not all models are available in all regions. Try us-central1 or check available regions.

📈 Roadmap

• [ ] Batch processing API
• [ ] Streaming responses
• [ ] Multi-modal prompts for Gemini
• [ ] Advanced video editing (VEO)
• [ ] Image editing capabilities (Imagen)
• [ ] Custom model fine-tuning
• [ ] Webhook notifications
• [ ] Usage analytics dashboard

🚀 Available on Smithery

⚠️ Note: Smithery deployment is currently experiencing technical issues. Please use the manual installation method in the Quickstart section above or see ALTERNATIVE_DEPLOYMENT.md for other options.

This MCP server is available on Smithery, the MCP server registry:

Quick Install via Smithery CLI

npx @smithery/cli install google-ai-mcp

Features on Smithery

• ✅ One-click installation
• ✅ Automatic Claude Desktop configuration
• ✅ Managed updates
• ✅ Community support
• ✅ Usage analytics

Visit our Smithery page for more information.

🤝 Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

📄 License

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

🙏 Acknowledgments

• Built with the Model Context Protocol SDK- Google Cloud Platform and Vertex AI teams
• Anthropic for creating the MCP protocol
• The open source community

📞 Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions
• Email: [email protected]

🌟 Star History

Made with ❤️ by Steve Kaplan