nanobanana-mcp

Overview

NanoBanana MCP provides a seamless integration between Google's Gemini image models and Claude applications. It enables users to generate and edit images directly within Claude environments, supporting both flash and pro models for different quality needs. The server maintains session consistency for character/style across multiple generations and allows runtime model switching without restarting the environment.

Try asking AI

After installing, here are 6 things you can ask your AI assistant:

• set_aspect_ratio

  Set aspect ratio before image generation

  Try:Set aspect ratio to 16:9Try:Use 1:1 square aspect ratio
• set_model

  Switch between flash/pro models at runtime

  Try:Switch to Pro model for higher qualityTry:Use Flash model for faster generation
• gemini_generate_image

  Generate images from text prompts

  Try:Create an image of a cyberpunk cityscape at sunsetTry:Generate a picture of a red-hat cat
• gemini_edit_image

  Edit images with natural language instructions

  Try:Edit the last image to change the hat color to blueTry:Add sunglasses to the person in the image
• gemini_chat

  Multi-turn conversation with images

  Try:Let's discuss the image you just generatedTry:Tell me more about the background in this image
• get_image_history

  View session image history

  Try:Show me all images from this sessionTry:List the previous images we've generated
• clear_conversation

  Reset session context

  Try:Clear the conversation historyTry:Start fresh with no previous context

Comparable tools

Installation

{
  "mcpServers": {
    "nanobanana-mcp": {
      "command": "npx",
      "args": ["-y", "@ycse/nanobanana-mcp"],
      "env": {
        "GOOGLE_AI_API_KEY": "your_api_key"
      }
    }
  }
}

claude mcp add nanobanana-mcp -- npx -y @ycse/nanobanana-mcp \
  -e "GOOGLE_AI_API_KEY=your_api_key"

FAQ

What models are supported?

The server supports Gemini Flash (gemini-3.1-flash-image-preview) and Gemini Pro (gemini-3-pro-image-preview) models. You can switch between them during a session without restarting.

How do I maintain character consistency across multiple image generations?

Use the same conversation_id parameter in your generation calls and set use_image_history to true. This maintains the visual context from previous images in the same conversation session.

Where are the generated images saved?

Images are saved to ~/Documents/nanobanana_generated/ by default. Generated files are named with a timestamp prefix, while edited images include the original filename.

Compare nanobanana-mcp with