Teaching Claude Code to Direct: A Stateful Video-Editing Skill Built on Gemini's Interactions API and MCP

Short summary

A developer wraps Google's Gemini Omni Flash video-generation model in a FastMCP server and packages it as a Claude Code skill, enabling stateful, multi-turn video editing from the terminal. The post explains how Gemini's Interactions API preserves visual context across turns, covering five generation modes (text-to-video, image animation, keyframe interpolation, subject-consistent generation, and video restyling). It also details practical constraints like synchronous generation, per-call billing, and large-output delivery via the File API.

• •MCP server exposes eight tools for stateful video generation and editing through Claude Code
• •Gemini Omni Flash's Interactions API keeps visual context server-side for iterative refinement
• •Covers five multimodal input modes and practical constraints like billing and file-size limits

Generated with AI, which can make mistakes.