Nano Banana MCP Server

A Model Context Protocol (MCP) server for AI-powered image generation using Google's Gemini 2.5 Flash Image Preview model. This server enables text-to-image and image-to-image generation through the MCP interface.

Features

• Text-to-Image Generation: Generate images from text prompts

• Image-to-Image Generation: Create variations or modifications of existing images

• Base64 Support: Handle image data in base64 format for MCP compatibility

• Google AI Studio Integration: Uses Google's latest Gemini 2.5 Flash Image Preview model

• Error Handling: Comprehensive error handling and logging

• Environment Variable Support: Secure API key management

Prerequisites

• Python 3.7+

• Google AI Studio API key

• Internet connection for API calls

Installation

1. Clone or download this repository

2. Install dependencies:

   pip install -r requirements.txt

3. Set up your Google AI Studio API key:

   export GOOGLE_AI_STUDIO_API_KEY="your_api_key_here"

Configuration

MCP Configuration

Add the following to your MCP configuration file:

{
  "mcpServers": {
    "nano-banana": {
      "command": "python3",
      "args": ["nano_banana_server.py"],
      "env": {
        "GOOGLE_AI_STUDIO_API_KEY": "${GOOGLE_AI_STUDIO_API_KEY}"
      }
    }
  }
}

Environment Variables

• GOOGLE_AI_STUDIO_API_KEY: Your Google AI Studio API key (required)

Usage

MCP Tool Function

The server provides a single MCP tool function:

generate_image_with_nano_banana

Generate images using Google's Gemini 2.5 Flash Image Preview model.

Parameters:

• prompt (string, required): Text description of the image to generate

• input_image_b64 (string, optional): Base64 encoded input image for variations

• input_image_mime_type (string, optional): MIME type of input image (e.g., 'image/png', 'image/jpeg')

Returns:

• success (boolean): Whether the generation was successful

• generated_image_b64 (string): Base64 encoded generated image (if successful)

• generated_image_size (integer): Size of generated image in bytes

• model_used (string): Name of the model used ("gemini-2.5-flash-image-preview")

• error (string): Error message (if failed)

Example Usage

Text-to-Image Generation

result = await generate_image_with_nano_banana(
    prompt="A beautiful handcrafted ceramic bowl with intricate patterns"
)

Image-to-Image Generation

result = await generate_image_with_nano_banana(
    prompt="Transform this into a watercolor painting",
    input_image_b64="iVBORw0KGgoAAAANSUhEUgAA...",  # Base64 encoded image
    input_image_mime_type="image/png"
)

Testing

Run the test function to verify the server is working:

python nano_banana_server.py

This will generate a test image and display the results.

API Reference

Classes

ImageGenerationRequest

Data class for image generation requests.

Attributes:

• prompt (str): Text prompt for image generation

• input_image_data (Optional[bytes]): Raw image data

• input_image_mime_type (Optional[str]): MIME type of input image

ImageGenerationResponse

Data class for image generation responses.

Attributes:

• success (bool): Whether generation was successful

• generated_image_data (Optional[bytes]): Generated image data

• error_message (Optional[str]): Error message if failed

• model_used (str): Model name used for generation

NanoBananaMCPServer

Main server class for handling image generation.

Methods:

• __init__(api_key: Optional[str] = None): Initialize server with API key

• generate_image(request: ImageGenerationRequest) -> ImageGenerationResponse: Generate image from request

• generate_image_from_base64(prompt: str, input_image_b64: Optional[str] = None, input_image_mime_type: Optional[str] = None) -> Dict[str, Any]: Generate image with base64 input