Nano Banana Pro MCP Server

How to use

You access this server from an MCP client and call its image tools to generate or edit images from prompts. The core capabilities you can rely on are: generate new images from text prompts, edit existing images with text instructions, continue editing the last image, compose multiple reference images into a new composition, and optionally ground image generation with real‑time information. You can specify output size and aspect ratios to fit your project, and you can enable real‑time search grounding when you need up‑to‑date visuals.

How to install

Prerequisites you must have before starting: Docker Desktop installed and running. You also need a Gemini API key from Google AI Studio to enable the image generation backend.

Step 1: Build the Docker image for the MCP server.

cd nano-banana-pro-mcp
docker build -t nano-banana-pro-mcp .

Step 2: Create the directories that will hold input and output files.

mkdir -p output input

Step 3: Add the MCP server configuration to Claude Code using one of the two options below.

Option A: Using claude mcp add (recommended). This registers the MCP server with Claude Code and uses a docker run invocation to start the server.

claude mcp add nano-banana-pro \
  --transport stdio \
  -- docker run -i --rm \
    -e GEMINI_API_KEY=$GEMINI_API_KEY \
    -v $(pwd)/output:/output \
    -v $(pwd)/input:/input:ro \
    nano-banana-pro-mcp

Option B: Manual configuration in Claude Code MCP config. This is a direct JSON configuration you place in your MCP settings.

{
  "mcpServers": {
    "nano-banana-pro": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "GEMINI_API_KEY",
        "-v", "$(pwd)/output:/output",
        "-v", "$(pwd)/input:/input:ro",
        "nano-banana-pro-mcp"
      ],
      "env": {
        "GEMINI_API_KEY": "your-api-key-here"
      }
    }
  }
}

Additional configuration and notes

The server exposes a set of image tools you can call from your MCP client. You can provide an output aspect ratio and resolution when generating or editing images, and you can enable Google Search grounding for real‑time information. Generated assets are stored in the /output directory by default, which is mounted from your host system according to how you start the container.

Troubleshooting

If the Gemini API key is not set properly, confirm that GEMINI_API_KEY is provided in the environment and that your shell exports it before starting the container. If images do not appear, verify that the /output directory is writable and correctly mounted. On Linux, you may need to adjust permissions for the output directory to allow the current user to write files.

Development and contributing

For local development, you can install dependencies, build, and run the server in a Node/npm environment if you prefer to run outside Docker. Ensure your GEMINI_API_KEY is available and set OUTPUT_DIR to the directory you want images saved to.

License and acknowledgments

MIT Licensed. Acknowledgments go to Google Gemini for the image generation API and to the Model Context Protocol ecosystem for enabling MCP integrations.

generate_image

Create a new image from a text prompt with optional aspect ratio and resolution settings.

edit_image

Modify an existing image using a text prompt, with optional reference images, aspect ratio, and resolution.

continue_editing

Iteratively refine the last generated image by supplying additional prompts and optional references.

compose_images

Combine multiple input images into a single composition guided by a prompt; supports up to 14 images.

get_last_image_info

Retrieve information about the most recently generated or edited image.

get_configuration_status

Check whether the API key and configuration are ready to use.