Gemini Nano Banana MCP

An MCP (Model Context Protocol) server for AI-powered image generation and editing using Google Gemini. Works with Claude Code, Cursor, and any MCP-compatible client.

Features

• Text-to-Image Generation - Generate images from text prompts via Gemini AI
• Image Editing - Edit existing images with natural language instructions
• Reference Images - Use reference images for style and content guidance
• Session Memory - Continue editing the last image without re-specifying the path
• Configurable Model - Choose any Gemini model (default: gemini-2.5-flash)
• Image History - Track and browse recently generated/edited images
• Cross-Platform - Works on macOS, Windows, and Linux

Quick Start

1. Get a Gemini API Key

Get your free API key from Google AI Studio.

2. Install

npm install -g @seungmanchoi/nano-banana-mcp

Or install from source:

git clone https://github.com/seungmanchoi/nano-banana-mcp.git
cd nano-banana-mcp
npm install
npm run build

3. Configure Your MCP Client

Claude Code

Add to ~/.claude/settings.json:

{
  "mcpServers": {
    "nano-banana": {
      "command": "npx",
      "args": ["-y", "@seungmanchoi/nano-banana-mcp"],
      "env": {
        "GEMINI_API_KEY": "your-api-key-here"
      }
    }
  }
}

Cursor

Add to your MCP settings:

{
  "mcpServers": {
    "nano-banana": {
      "command": "npx",
      "args": ["-y", "@seungmanchoi/nano-banana-mcp"],
      "env": {
        "GEMINI_API_KEY": "your-api-key-here"
      }
    }
  }
}

From Source

If installed from source, use the absolute path:

{
  "mcpServers": {
    "nano-banana": {
      "command": "node",
      "args": ["/absolute/path/to/nano-banana-mcp/dist/index.js"],
      "env": {
        "GEMINI_API_KEY": "your-api-key-here"
      }
    }
  }
}

You can also skip the env field and configure the API key at runtime using the configure_api_key tool.

Model Configuration

The default model is gemini-2.5-flash. You can change it in several ways:

Option 1: Environment Variable

Set GEMINI_MODEL in your MCP client config:

{
  "mcpServers": {
    "nano-banana": {
      "command": "npx",
      "args": ["-y", "@seungmanchoi/nano-banana-mcp"],
      "env": {
        "GEMINI_API_KEY": "your-api-key-here",
        "GEMINI_MODEL": "gemini-2.0-flash-exp"
      }
    }
  }
}

Option 2: Runtime Tool

Use the configure_model tool to change the default model at runtime. The setting persists across sessions in ~/.nano-banana/config.json.

Set the model to gemini-2.5-pro

Option 3: Per-Request Override

Pass the model parameter directly to generate_image, edit_image, or continue_editing to override the default for a single request:

Generate an image of a cat using model gemini-2.0-flash-exp

Model Priority

1. Per-request model parameter (highest priority)
2. GEMINI_MODEL environment variable
3. Config file (~/.nano-banana/config.json)
4. Default: gemini-2.5-flash

Available Models

For the latest list of models that support image generation, see Google AI documentation.

Tools

Usage Examples

Once the MCP server is connected, you can use natural language in your MCP client:

Generate an image of a sunset over mountains with a lake reflection

Edit the image at ~/nano-banana-images/gen_2025-01-01.png to add a boat on the lake

Continue editing - make the sky more vibrant with orange and pink tones

Show me the last 5 images I generated

Switch to gemini-2.5-pro model for higher quality

API Key Configuration

The server loads the API key in the following priority order:

1. Environment variable - GEMINI_API_KEY
2. Config file - ~/.nano-banana/config.json
3. Runtime - via the configure_api_key tool

Image Storage

Generated and edited images are automatically saved to:

Project Structure

src/
├── index.ts              # Entry point
├── server.ts             # MCP server setup and request routing
├── config/
│   └── settings.ts       # API key and model management
├── services/
│   ├── gemini.ts         # Google Gemini API client
│   └── storage.ts        # Image file I/O and history tracking
├── tools/
│   ├── definitions.ts    # MCP tool schemas
│   └── handlers.ts       # Tool request handlers
└── types/
    └── index.ts          # TypeScript type definitions

Development

npm run dev          # Run with tsx (no build needed)
npm run build        # Compile TypeScript
npm run typecheck    # Type check without emitting
npm run lint         # Run ESLint

Tech Stack

• Runtime: Node.js
• Language: TypeScript (strict mode, ES2022)
• MCP SDK: @modelcontextprotocol/sdk
• AI Model: Google Gemini (configurable, default: gemini-2.5-flash)
• Validation: Zod

License

MIT