Teaching Codex to Paint: A Stateful Image-Editing Skill Built on Gemini's Interactions API and MCP

Short summary

nb2lite-skill-codex wraps Google's gemini-3.1-flash-lite-image model in a FastMCP server packaged as a Codex skill, enabling stateful image generation and iterative editing within a coding session. Using Gemini's Interactions API, the tool preserves visual context across turns so you can say 'add a neon sign' without re-describing the whole scene. The MCP server exposes four tools: generate, edit, edit-local, and list interactions.

• •Wraps Gemini's NB2Lite image model in a FastMCP server as a Codex skill for stateful image editing
• •Uses Gemini Interactions API to preserve visual context across multiple editing turns
• •Exposes four MCP tools: generate_image, edit_image, edit_local_image, and list interactions

Generated with AI, which can make mistakes.