Connects Claude to Google's Gemini image generation API with two tools: generate_image for text-to-image, multi-turn editing with reference images, and iterative refinement via session IDs, plus process_image for local operations like cropping, background removal, chroma keying, and format conversion. Built on Gemini's current generateContent API rather than the deprecated Imagen service. Includes built-in cost tracking per request and session, configurable rate limits to prevent runaway agent spending, and auto-discovery of available models from your API key. Saves outputs with auto-versioning, maintains a generation manifest in JSONL, and supports aspect ratios from 1:1 to 21:9 at 1K/2K/4K resolutions. Good fit if you need programmatic image generation with cost controls or want to batch process images locally without hitting APIs.
claude mcp add --transport stdio io.github.jimothysnicket-gemini-image uvx gemini-image