mcp-nano-banana

mcp-nano-banana

MCP server for AI-powered image generation using Google Gemini Nano Banana models. Generate, edit, and combine images from text prompts — directly from your AI coding assistant.

Features

• Text-to-image generation — Create images from natural language descriptions
• Image editing — Upload images and modify them with text instructions
• Multi-image combination — Mix up to 14 reference images into one composite
• Resolution control — Generate at 512px, 1K, 2K, or 4K resolution
• 13+ aspect ratios — From square (1:1) to ultra-wide (21:9) and tall (1:8)
• Auto-fallback — Pro model falls back to Flash on rate limits (429/403/402)
• File output — Save to specific paths or return inline base64
• Favicon generation — Generate complete favicon packages with one command
• Multi-format output — PNG, JPEG, and WebP support

Quick Start

Prerequisites

1. Get a Gemini API key from Google AI Studio
2. Enable billing on your Google Cloud project — image generation requires a paid tier (~$0.04-0.05/image for NB2 at standard resolution)

Installation

npm install -g mcp-nano-banana

Set your API key:

export GEMINI_API_KEY=your-api-key

For MCP Registry publication metadata, the registry name is io.github.codeFi/mcp-nano-banana.

MCP Configuration

Codex

codex mcp add nano-banana --env GEMINI_API_KEY=your-api-key -- npx -y mcp-nano-banana@latest

Or add manually to ~/.codex/config.toml:

[mcp_servers.nano-banana]
command = "npx"
args = ["-y", "mcp-nano-banana@latest"]

[mcp_servers.nano-banana.env]
GEMINI_API_KEY = "your-api-key"

Qwen Code

Add to ~/.qwen/settings.json:

{
  "mcpServers": {
    "nano-banana": {
      "command": "npx",
      "args": ["-y", "mcp-nano-banana@latest"],
      "env": {
        "GEMINI_API_KEY": "your-api-key"
      }
    }
  }
}

Claude Code

claude mcp add nano-banana -e GEMINI_API_KEY=your-api-key -- npx -y mcp-nano-banana@latest

Or add manually to ~/.claude/claude_desktop_config.json:

{
  "mcpServers": {
    "nano-banana": {
      "command": "npx",
      "args": ["-y", "mcp-nano-banana@latest"],
      "env": {
        "GEMINI_API_KEY": "your-api-key"
      }
    }
  }
}

Other MCP Clients

For CLIs and desktop clients that use the standard mcpServers JSON format, register:

{
  "mcpServers": {
    "nano-banana": {
      "command": "npx",
      "args": ["-y", "mcp-nano-banana@latest"],
      "env": {
        "GEMINI_API_KEY": "your-api-key"
      }
    }
  }
}

Local / Development Build

git clone https://github.com/codeFi/mcp-nano-banana.git
cd mcp-nano-banana
npm install
npm run build

Then register with your MCP client:

{
  "mcpServers": {
    "nano-banana": {
      "command": "node",
      "args": ["/path/to/mcp-nano-banana/dist/index.js"],
      "env": {
        "GEMINI_API_KEY": "your-api-key"
      }
    }
  }
}

Models

Model	MCP Name	Description	Max Ref Images
gemini-3-pro-image-preview	Nano Banana Pro	High quality, thinking mode, professional assets	6 object + 5 character
gemini-3.1-flash-image-preview	Nano Banana 2	Fast, high-efficiency, 1:4/4:1/1:8/8:1 ratios	10 object + 4 character
gemini-2.5-flash-image	Nano Banana	Fastest, optimized for high-volume tasks	-

The server defaults to Nano Banana Pro for best quality and automatically falls back to Flash on rate limits.

Available Parameters

Resolution

Value	Approx. Pixels	Supported Models
"512"	~512px	Nano Banana 2 only
"1K"	~1024px	All models
"2K"	~2048px	All models
"4K"	~4096px	All models

Default: "1K"

Aspect Ratio

Value	Dimensions at 1K	Common Use Case
1:1	1024×1024	Social posts, icons
16:9	1344×768	Headers, presentations
9:16	768×1344	Stories, mobile
4:3	1152×896	Desktop, photos
3:4	896×1152	Portraits
2:3	~896×1344	Print, magazines
3:2	~1344×896	Photos
4:5	~896×1120	Instagram posts
5:4	~1120×896	Instagram landscape
21:9	~2048×878	Ultra-wide banners
1:4	~512×2048	Nano Banana 2 only — tall strips
4:1	~2048×512	Nano Banana 2 only — wide strips
1:8	~512×4096	Nano Banana 2 only — extra tall
8:1	~4096×512	Nano Banana 2 only — extra wide

Tools

generate_image

Generate or edit images from text prompts.

Parameter	Type	Required	Default	Description
prompt	string	✅ Yes	—	Text description or editing instructions
imagePaths	string[]	No	—	Reference image paths on disk (up to 14)
model	string	No	gemini-3-pro-image-preview	Model to use
aspectRatio	string	No	1:1	See table above for all options
resolution	string	No	1K	512, 1K, 2K, or 4K
outputFormat	string	No	png	png, jpeg, webp
outputPath	string	No	—	Save image to this file path

Examples

Generate from text:

{
  "prompt": "A sunset over mountains with dramatic lighting",
  "aspectRatio": "16:9",
  "resolution": "2K",
  "outputPath": "/Users/currentUser/Desktop/sunset.png"
}

Edit an existing image:

{
  "prompt": "Change the background to a coffee shop interior",
  "imagePaths": ["/Users/currentUser/photo.jpg"],
  "outputPath": "/Users/currentUser/edited.png"
}

Combine multiple images:

{
  "prompt": "Put these glasses on this person's face",
  "imagePaths": ["/path/to/face.jpg", "/path/to/glasses.jpg"],
  "outputPath": "/path/to/result.png"
}

High-res wallpaper:

{
  "prompt": "A futuristic cityscape at night with neon lights and rain",
  "aspectRatio": "16:9",
  "resolution": "4K",
  "outputPath": "/path/to/wallpaper.png"
}

generate_favicons

Generate a complete favicon package from a prompt or existing image.

Parameter	Type	Required	Default	Description
prompt	string	No*	—	Icon description (required if imagePath not set)
imagePath	string	No*	—	Path to existing master icon image
model	string	No	gemini-3-pro-image-preview	Model for prompt generation
outputDir	string	No	—	Directory for individual favicon files
outputPath	string	No	—	Save ZIP package to this path
returnZipBase64	boolean	No	false	Return ZIP as base64 text
appName	string	No	My App	Manifest name
shortName	string	No	App	Manifest short_name
themeColor	string	No	#ffffff	Manifest theme color
backgroundColor	string	No	#ffffff	Manifest background color
includeReadme	boolean	No	true	Include README.md in package

Example

{
  "prompt": "A minimalist banana icon with a soft gradient",
  "outputDir": "/path/to/favicon",
  "outputPath": "/path/to/favicon.zip",
  "appName": "MyBrand",
  "shortName": "Brand",
  "themeColor": "#1a1a2e",
  "backgroundColor": "#ffffff"
}

Generated files:

• favicon.ico — Multi-size icon (16/32/48px)
• favicon-16x16.png — Browser tabs
• favicon-32x32.png — Higher-res tabs
• apple-touch-icon.png — iOS home screen
• android-chrome-192x192.png — Android
• android-chrome-512x512.png — Android HD
• site.webmanifest — PWA manifest

HTML snippet (included in response):

<link rel="icon" href="/favicon.ico" sizes="any">
<link rel="apple-touch-icon" sizes="180x180" href="/apple-touch-icon.png">
<link rel="icon" type="image/png" sizes="32x32" href="/favicon-32x32.png">
<link rel="icon" type="image/png" sizes="16x16" href="/favicon-16x16.png">
<link rel="manifest" href="/site.webmanifest">

Pricing

Image generation requires a paid Google AI Studio account. Approximate costs:

Model	1K	2K	4K
Nano Banana 2	~$0.04	~$0.08	~$0.15
Nano Banana Pro	~$0.06	~$0.13	~$0.24

You only pay for what you generate — no monthly minimum.

Supported Image Formats

Input (reference images): png, jpg, jpeg, webp, gif Output: png, jpeg, webp

Development

# Build TypeScript
npm run build

# Watch mode
npm run dev

# Run tests
npm test

# Run the server directly
node dist/index.js

Troubleshooting

Issue	Solution
GEMINI_API_KEY is not set	Export the env var or add it to your MCP config
429 Too Many Requests	Server auto-falls back to Flash model. Add billing to your account.
403/402 Quota exceeded	Enable billing at Google Cloud Console
503 Unavailable	High demand — retry in a moment. Common with 4K requests.
Image not updating after edit	Make sure you rebuilt the TypeScript (npm run build)
imagePaths not found	Use absolute paths. Relative paths may not resolve correctly.

Architecture

AI Assistant (MCP Client)
       │  (MCP Protocol over stdio)
       ▼
mcp-nano-banana Server
       │  (HTTPS POST)
       ▼
Google Gemini API
       │
       ▼
Generated Image → saved to outputPath (or returned inline base64)

License

MIT

Author

codeFi