Supercharge Claude Desktop and Claude Code with Google's Gemini multimodal capabilities! Generate stunning images with session-based consistency, edit existing ones, and leverage advanced vision AI - all within your Claude environment.

• 🎨 Image Generation - Create 2K images from text prompts using Gemini 3 Pro
• 🔄 Image Consistency - Maintain character/style consistency across multiple generations within a session
• 🖼️ Image Editing - Transform existing images with natural language instructions
• 🔍 Google Search Integration - Ground image generation with real-world references
• 💬 Multi-turn Chat - Maintain conversational context across interactions
• 📜 Session History - Reference previous images using last or history:N

# Generate an image
"Create a serene Korean beach scene with traditional architecture"

# Edit an existing image
"Add a dramatic T-Rex appearing on the beach, people reacting with surprise"

• Node.js 18+
• One of: Claude Desktop, Claude Code, VSCode, Cursor, or Windsurf
• Google AI API Key (Get it here)

First, clone and build the project:

git clone https://github.com/YCSE/nanobanana-mcp.git
cd nanobanana-mcp
npm install
npm run build

# Configure API Key
cp .env.example .env
# Edit .env and add your GOOGLE_AI_API_KEY

Then choose your platform:

Edit your Claude Desktop config:

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

{
  "mcpServers": {
    "nanobanana-mcp": {
      "command": "node",
      "args": ["/absolute/path/to/nanobanana-mcp/dist/index.js"],
      "env": {
        "GOOGLE_AI_API_KEY": "your_api_key_here"
      }
    }
  }
}

Restart Claude Desktop after adding the configuration.

# After building, install to Claude Code
source .env && claude mcp add nanobanana-mcp "node" "dist/index.js" \
  -e "GOOGLE_AI_API_KEY=$GOOGLE_AI_API_KEY"

Install the Continue extension and add to ~/.continue/config.json:

{
  "models": [
    // Your existing models
  ],
  "mcpServers": {
    "nanobanana-mcp": {
      "command": "node",
      "args": ["/absolute/path/to/nanobanana-mcp/dist/index.js"],
      "env": {
        "GOOGLE_AI_API_KEY": "your_api_key_here"
      }
    }
  }
}

Add to your Cursor settings file ~/.cursor/config.json:

{
  "mcpServers": {
    "nanobanana-mcp": {
      "command": "node",
      "args": ["/absolute/path/to/nanobanana-mcp/dist/index.js"],
      "env": {
        "GOOGLE_AI_API_KEY": "your_api_key_here"
      }
    }
  }
}

Add to your Windsurf configuration file ~/.windsurf/config.json:

{
  "mcpServers": {
    "nanobanana-mcp": {
      "command": "node",
      "args": ["/absolute/path/to/nanobanana-mcp/dist/index.js"],
      "env": {
        "GOOGLE_AI_API_KEY": "your_api_key_here"
      }
    }
  }
}

Generate 2K images from text descriptions with session-based consistency.

{
  prompt: string;              // Image description
  output_path?: string;        // Optional save path (default: ~/Documents/nanobanana_generated/)
  conversation_id?: string;    // Session ID for image history
  use_image_history?: boolean; // Use previous images for style/character consistency
  reference_images?: string[]; // Manual reference images for consistency
  enable_google_search?: boolean; // Enable Google Search for real-world grounding
}

Example - Basic:

"Generate a cyberpunk cityscape at sunset with flying cars"

Example - With Consistency:

// First image
{ prompt: "A cute red-hat cat", conversation_id: "cat-session" }

// Second image - maintains same character
{ prompt: "The same cat taking a nap", conversation_id: "cat-session", use_image_history: true }

Edit existing images using natural language. Supports session history references.

{
  image_path: string;          // Path, or "last", or "history:N"
  edit_prompt: string;         // Edit instructions
  output_path?: string;        // Optional save path
  conversation_id?: string;    // Session ID for accessing history
  reference_images?: string[]; // Additional style references
  enable_google_search?: boolean; // Enable Google Search
}

Example - File Path:

"Remove the background and make it transparent"

Example - History Reference:

// Edit the most recent image in the session
{ image_path: "last", edit_prompt: "Change hat color to blue", conversation_id: "cat-session" }

// Edit a specific image from history
{ image_path: "history:0", edit_prompt: "Add sunglasses", conversation_id: "cat-session" }

Chat with Gemini for general queries.

{
  message: string;           // Your message
  conversation_id?: string;  // Optional conversation ID
  system_prompt?: string;    // Optional system instructions
}

View generated/edited images in a session.

{
  conversation_id: string;   // Session to view
}

Response includes:

• Image index and reference (history:0, history:1, etc.)
• File paths
• Original prompts
• Timestamps

Reset conversation history.

{
  conversation_id: string;   // Conversation to clear
}

NanoBanana MCP uses different Gemini models optimized for each task:

To switch between model variants (e.g., nanobanana vs nanobanana-pro), edit the model constants in src/index.ts:

// For image generation (line ~428)
const model = 'gemini-3-pro-image-preview';  // Pro version (current)
// const model = 'gemini-2.5-flash-image-preview';  // Flash version (faster, lower quality)

// For image editing (line ~659)
const model = 'gemini-3-pro-image-preview';  // Pro version (current)
// const model = 'gemini-2.5-flash-image-preview';  // Flash version

After editing, rebuild the project:

npm run build

Model Comparison:

• Generate placeholder images for web development
• Create app icons and assets
• Analyze UI/UX screenshots
• Generate test data images

• Edit images with text commands
• Generate blog illustrations
• Create social media visuals
• Batch process image modifications

• Rapid prototyping with generated visuals
• Style transfer and variations
• Color scheme analysis
• Accessibility checking

Images are automatically saved to:

• Generated images: ~/Documents/nanobanana_generated/generated_[timestamp].png
• Edited images: ~/Documents/nanobanana_generated/[original_name]_edited_[timestamp].png

All images are saved in PNG format for maximum quality.

# Run in development mode with hot reload
npm run dev

# Build for production
npm run build

# Type checking
npm run typecheck

nanobanana-mcp/
├── src/
│   └── index.ts         # MCP server implementation
├── dist/                # Compiled JavaScript
├── .env                 # API configuration
├── claude-mcp           # CLI management tool
└── package.json

• API keys are stored locally in .env
• Never commit .env to version control
• All image operations happen locally
• No data is stored on external servers

# Check installation
./claude-mcp status

# Rebuild if needed
npm run build

• Verify API key is valid
• Check API quota at Google AI Studio
• Ensure output directory has write permissions

1. Restart Claude Desktop/Code
2. Check config file syntax
3. Verify absolute paths in configuration

Contributions are welcome! Please feel free to submit a Pull Request.

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

MIT License - see LICENSE file for details.

If you find this project useful, please consider giving it a star ⭐️

• Anthropic for Claude and MCP
• Google for Gemini API
• Model Context Protocol community

• 🐛 Report Issues
• 💬 Discussions
• 📖 Documentation

Made with ❤️ for the Claude community