Connects Claude to image generation and editing through Gemini and OpenAI APIs. Exposes generate_image, edit_image, and edit_last tools that let you iterate on results conversationally without re-specifying paths. Each generation starts a session with undo/redo support, so you can step backward through edits or branch off from any point. Includes a Claude Code skill that teaches the agent 24 editing techniques and platform-specific sizing rules for social media and app stores. History and output land in .imgx/ at your project root. Reach for this when you need to generate mockups, iterate on designs, or produce visual assets without leaving your coding environment.
claude mcp add --transport stdio io.github.somacoffeekyoto-imgx -- npx -y imgx-mcp