Nano Banana Pro MCP

# CLAUDE.md This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository. ## Project Overview MCP (Model Context Protocol) server that wraps Google's Gemini 3 Nano Banana Pro image generation API, enabling AI agents to generate images via a standardized tool interface. ## Commands ```bash npm run build # Compile TypeScript to dist/ npm test # Run tests with vitest npm run test:watch # Run tests in watch mode npm run typecheck # Type check without emitting npm start # Run the MCP server (requires GEMINI_API_KEY env var) ``` ## Architecture - `src/index.ts` - MCP server setup with tool registration. Exports `createServer(apiKey)` for testing. - `src/gemini.ts` - `GeminiImageClient` class that handles Gemini API requests - `src/types.ts` - TypeScript interfaces for Gemini API request/response shapes The server exposes one tool: `generate_image` with parameters for prompt, aspect ratio, and image size. ## Testing Tests use vitest with mocked fetch. The MCP server tests use `@modelcontextprotocol/sdk`'s `InMemoryTransport` for end-to-end tool invocation testing without network calls. ## Important - Never commit API keys. The `.gitignore` excludes `.env` files. - The Gemini API returns images as base64-encoded data which the MCP server passes through as `image` content type.