nanobanana

NanoBanana MCP exposes Gemini-based image generation and editing capabilities within Claude Desktop, Claude Code, and Cursor. It supports switching between Nano Banana 2 (Flash) and Nano Banana Pro models at runtime so you can iterate quickly or render higher-quality outputs as needed. Core tools include gemini_generate_image for creating images from prompts, gemini_edit_image for applying natural-language edits to existing images, and gemini_chat for multi-turn conversations with image support. Maintain session continuity with set_aspect_ratio to ensure consistent framing, and use get_image_history and clear_conversation to manage evolving visuals across a session.

Prerequisites:

• Node.js 18+ and npm
• Access key for Google AI API (GOOGLE_AI_API_KEY)

Installation steps:

1. Clone the MCP server repository or install the published package: git clone https://github.com/YCSE/nanobanana-mcp.git cd nanobanana-mcp

2. Install dependencies: npm install

3. Set up environment variables (example):

   • GOOGLE_AI_API_KEY=your_api_key
   • NANOBANANA_MODEL=gemini-3.1-flash-image-preview or gemini-3-pro-image-preview

4. Run in development or start mode: npm run dev # Development mode with hot reload npm run build # Build production assets npm run start # Run the compiled server

5. Add to MCP configuration in your environment as shown in the mcp_config section.

Environment variables

• GOOGLE_AI_API_KEY is required for image generation and editing
• NANOBANANA_MODEL can be used to lock a default model; you can switch models at runtime via the provided tools Common issues
• If tools don’t appear, restart Claude Desktop/Code and verify the MCP config syntax
• Ensure the npx command runs without errors: npx -y @ycse/nanobanana-mcp Configuration tips
• Use set_aspect_ratio before generation/edit to ensure correct framing
• Use gemini_chat for context-aware multi-turn conversations with image support
• Reference images and history can help maintain character consistency across renders