Nano Banana Pro MCP
Enables Claude to generate and edit images using Google's Gemini image models, with options for resolution and batch processing, and optional Firebase Storage for cross-environment image transfer.
README
Nano Banana Pro MCP Server
A minimal, auditable MCP server that gives Claude Desktop the ability to generate and edit images using Google's Nano Banana Pro (Gemini 3 Pro Image) model.
~150 lines of TypeScript. No hidden dependencies. You own the code.
What it does
| Tool | Description |
|---|---|
set_model |
Switch between Gemini image models |
set_resolution |
Set source resolution (1K/2K/4K) |
generate_image |
Create an image from a text prompt |
generate_image_batch |
Create several images at once |
edit_image |
Edit an image (given its URL) with instructions |
Generated images are returned inline (so Claude can display them in chat) and saved to ~/nano-banana-output/.
Images are never written to local disk. Each is returned inline for viewing, and — when Firebase Storage is configured — uploaded there with a signed URL you can fetch from other environments (e.g. to embed in a document).
Prerequisites
- Node.js 18+ — check with
node --version - Gemini API key — free from Google AI Studio. Enable billing for image generation (pay-per-image, no monthly minimum).
Setup
1. Clone and build
git clone <your-repo-url> nano-banana-mcp
cd nano-banana-mcp
npm install
npm run build
2. Configure Claude Desktop
Open your Claude Desktop config file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
Add this entry (adjust the path to where you cloned the repo):
{
"mcpServers": {
"nano-banana-pro": {
"command": "node",
"args": ["/absolute/path/to/nano-banana-mcp/build/index.js"],
"env": {
"GEMINI_API_KEY": "your-api-key-here"
}
}
}
}
3. Restart Claude Desktop
The server should appear in the MCP tools menu (hammer icon).
Usage examples
Once connected, just ask Claude naturally:
- "Generate an image of a Senegalese market at sunset"
- "Create a 16:9 hero banner for a blog about climate resilience"
- "Edit ~/nano-banana-output/gen_20260714_a1b2.png — remove the background and add a gradient"
Models
The server supports two Gemini image models:
| Model | ID | Quality | Speed | Cost |
|---|---|---|---|---|
| Nano Banana Pro | gemini-3-pro-image-preview |
Highest | Slower | ~$0.09/image |
| Nano Banana | gemini-2.5-flash-image |
Good | Fast | ~$0.04/image |
By default, generate_image uses Pro. Pass use_pro: false to switch to the faster Flash model.
Security notes
- Your API key stays local — it's passed via environment variable, never logged or transmitted elsewhere.
- The server only makes HTTPS calls to
generativelanguage.googleapis.com(Google's Gemini API). - No telemetry, no analytics, no third-party calls.
- All code is in
src/index.ts— read it before you run it.
Project structure
nano-banana-mcp/
├── src/
│ └── index.ts ← the entire server (~150 lines)
├── package.json
├── tsconfig.json
└── README.md
License
MIT — do whatever you want with it.
Optional: Firebase Storage (cross-environment image transfer)
By default, generated images are saved to ~/nano-banana-output/ on the machine running the MCP server. If you want images to be usable in another environment (e.g. embedding in a document built by Claude's sandbox) without manually attaching files, configure Firebase Storage. The server will upload each image and return a time-limited signed URL that the other environment can download directly.
Point at your service-account JSON key file (recommended):
"env": {
"GEMINI_API_KEY": "[API_KEY]",
"SERVICE_ACCOUNT_KEY_PATH": "/absolute/path/to/firebase-service-account.json",
"FIREBASE_STORAGE_BUCKET": "your-project.appspot.com"
}
Download the JSON key from Firebase Console → Project Settings → Service Accounts → Generate new private key, save it somewhere on your machine, and point SERVICE_ACCOUNT_KEY_PATH at it.
Alternatively, supply the individual fields instead of a key file (FIREBASE_PROJECT_ID, FIREBASE_CLIENT_EMAIL, FIREBASE_PRIVATE_KEY, plus FIREBASE_STORAGE_BUCKET).
Optional: FIREBASE_SIGNED_URL_MINUTES (default 60) controls how long the download URL stays valid.
Notes:
- The service-account credentials stay in the MCP server process only — they are never sent to Claude or its sandbox. The sandbox only ever sees a plain signed URL.
- Get the service-account values from Firebase Console → Project Settings → Service Accounts → Generate new private key.
- If any of the four vars is missing, uploads stay disabled and the server falls back to saving locally.
- For the download to work, the environment fetching the URL must have outbound network access to
storage.googleapis.com.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.