google-ai-mcp-server

Google AI MCP Server

A Model Context Protocol (MCP) server that provides access to Google's AI models including Imagen 3, VEO 3, Gemini, and Lyria through Claude Desktop.

🚀 Quick Start

Install via NPM (Recommended)

Add to your Claude Desktop configuration:

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

✨ Features

• Imagen 3 - Generate photorealistic images with text prompts
• VEO 3 - Create 5-8 second videos with audio (coming soon)
• Gemini - Access Google's latest language models
• Lyria 2 - Generate up to 60 seconds of music (coming soon)

📋 Prerequisites

1. Google Cloud Project with billing enabled
2. Vertex AI API enabled
3. Service account with appropriate permissions
4. Claude Desktop installed

🔧 Setup Instructions

1. Enable Required APIs

gcloud services enable aiplatform.googleapis.com

2. Create Service Account

# Create service account
gcloud iam service-accounts create google-ai-mcp-server \
    --display-name="Google AI MCP Server Service Account"

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

# Create and download key
gcloud iam service-accounts keys create service-account-key.json \
    --iam-account=google-ai-mcp-server@YOUR_PROJECT_ID.iam.gserviceaccount.com

3. Configure Claude Desktop

Update your Claude Desktop config file:

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

{
  "mcpServers": {
    "google-ai": {
      "command": "npx",
      "args": ["@stevekaplanai/google-ai-mcp@latest"],
      "env": {
        "USE_MOCK": "false",
        "GOOGLE_CLOUD_PROJECT": "your-project-id",
        "GOOGLE_CLOUD_LOCATION": "us-central1",
        "GOOGLE_APPLICATION_CREDENTIALS_JSON": "paste-json-here"
      }
    }
  }
}

⚠️ Important: When pasting the JSON credentials:

• Escape all newlines in the private key: \n → \\n
• Keep the entire JSON on one line
• Don't include outer quotes

4. Restart Claude Desktop

After updating the configuration, completely restart Claude Desktop for changes to take effect.

🎨 Usage Examples

Generate an Image

"Create a photorealistic image of a ninja duck in a dojo"

Generate Multiple Variations

"Generate 4 variations of a cyberpunk city at night"

Different Aspect Ratios

"Create a 16:9 landscape image of mountains at sunset"
"Generate a 9:16 portrait image for a phone wallpaper"

🛠️ Available Tools

imagen_generate_image

Generate images using Google Imagen 3.

Parameters:

• prompt (required): Text description of the image
• aspectRatio: "1:1" (default), "16:9", "9:16", "4:3", "3:4"
• sampleCount: 1-8 images (default: 1)
• negativePrompt: What to avoid in the generation
• personGeneration: "allow" (default) or "disallow"

gemini_generate_text

Generate text using Gemini models.

Parameters:

• prompt (required): Input text prompt
• model: "gemini-1.5-flash" (default), "gemini-1.5-pro", "gemini-2.0-flash-exp"
• temperature: 0-2 (default: 0.7)
• maxTokens: 1-8192 (default: 2048)

veo_generate_video (Coming Soon)

Generate videos with VEO 3.

lyria_generate_music (Coming Soon)

Generate music with Lyria 2.

🧪 Mock Mode

For testing without API calls, set USE_MOCK: "true" in your configuration. This will return sample responses without using your Google Cloud quota.

🔍 Troubleshooting

JSON Parsing Errors

If you see "Unexpected token" errors:

1. Update to the latest version: @stevekaplanai/google-ai-mcp@latest
2. Restart Claude Desktop

Authentication Errors

• Verify your service account has the aiplatform.user role
• Check that the JSON credentials are properly escaped
• Ensure the project ID matches your service account

API Not Found Errors

• Confirm Vertex AI API is enabled in your project
• Verify you're using a supported location (us-central1 recommended)

📊 Supported Models

Imagen

• imagen-3.0-generate-001 - Latest high-quality model
• imagen-3.0-fast-generate-001 - Faster generation

Gemini

• gemini-1.5-pro - Most capable model
• gemini-1.5-flash - Optimized for speed
• gemini-2.0-flash-exp - Experimental features

🔒 Security Notes

• Never commit service account keys to version control
• Use environment variables for production deployments
• Regularly rotate service account keys
• Limit service account permissions to minimum required

📝 Changelog

v1.0.2 (Latest)

• Fixed JSON parsing errors by redirecting debug logs to stderr
• Improved error handling for API responses
• Added support for real Imagen API endpoints

v1.0.1

• Initial release with mock mode
• Basic Imagen and Gemini support

🤝 Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

📄 License

MIT License - see LICENSE for details.

🙏 Acknowledgments

• Built for Claude Desktop by Anthropic
• Uses Google's Vertex AI platform
• Implements the Model Context Protocol

📧 Support

For issues and questions:

• GitHub Issues: google-ai-mcp-server/issues
• NPM Package: @stevekaplanai/google-ai-mcp

---

Made with ❤️ by Steve Kaplan