Image Generation MCP

# CLAUDE.md This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository. ## Commands ```bash npm run build # Compile TypeScript to dist/ npm run dev # Watch mode compilation npm test # Run all tests npm run test:watch # Run tests in watch mode npx vitest run tests/security.test.ts # Run a single test file npm run lint # Check for lint errors npm run lint:fix # Auto-fix lint errors npm run format # Format code with Prettier npm run check # Run all checks (typecheck + lint + format + test) ``` ## Architecture This is an MCP (Model Context Protocol) server that exposes image generation as tools for AI assistants like Claude Desktop. ### Provider Pattern The server uses a provider abstraction to support multiple image generation APIs: ``` src/providers/types.ts → ImageProvider interface src/providers/gemini.ts → Gemini/Nano Banana implementation src/providers/index.ts → Provider factory and registry ``` To add a new provider: implement `ImageProvider` interface, then register in `providers/index.ts`. ### MCP Tool Flow 1. `src/index.ts` - MCP server setup, registers handlers for `ListToolsRequestSchema` and `CallToolRequestSchema` 2. `src/tools/generate-image.ts` - Tool schema (Zod), execution logic, calls provider 3. `src/config/presets.ts` - Platform dimension presets (ghost-banner, instagram-post, etc.) ### Key Types - `ImageProvider` - Interface for generation backends - `PlatformPreset` - Dimension/aspect ratio config for platforms - `GenerateImageInput` - Zod schema for tool input validation - `ProviderError` - Typed error with retry info ## Environment Requires `GOOGLE_API_KEY` environment variable for Gemini provider. Store in `.env` file (gitignored). ## Testing Tests are in `tests/` directory using Vitest. The `gemini.integration.test.ts` makes live API calls when `GOOGLE_API_KEY` is set.