Nanobanana

You can run the Nanobanana MCP Server to generate images using Google Gemini models and manage generated outputs. This server exposes practical image generation capabilities and a simple way to list generated images, all through MCP-compatible commands. It is designed to be used with an MCP client that can send the generate_image and list_images requests and handle the resulting image files.

How to use

To use this MCP server, you provide an API key for Google Gemini image generation and start the server locally through MCP tooling. Once running, you can request image generation with the generate_image tool or list previously generated images with list_images. The server respects parameters such as prompt, model, aspect ratio, and output directory, helping you tailor results to your needs.

How to install

Prerequisites you need before installation:

• • Node.js 14+ recommended
• • npm present on your system

Install and run the MCP server using the following steps. Copy each command exactly as shown.

claude mcp add nanobanana -e GEMINI_API_KEY=your-key-here -- npx -y @saroby/nanobanana-mcp

Start the server locally

Run the MCP server using npm in dev/watch mode. This starts the server so you can begin issuing MCP calls to generate images or list existing ones.

npm install
npm run build
npm run dev  # watch mode

Available tools

generate_image

Generate images using Google Gemini models with options for prompt, model mode, aspect_ratio, negative_prompt, output_dir, and count.

list_images

List generated images in a specified directory, defaulting to ./nanobanana-images.