Sesori Review

Claude reviews your Figma prototype with you, inside Figma.
It walks the canvas screen by screen, asks before it assumes, and leaves Dev Mode annotations your developers can build from.

Why you'd want this

• A second pair of eyes before handoff. Click Review flow and Claude reads the prototype, then walks every screen: dead ends, missing loading/empty/error states, inconsistent spacing, unclear interactions, accessibility gaps.
• It moves your canvas. Claude focuses each frame before talking about it, so you follow along instead of reading a report about frames you have to hunt for.
• It asks, it doesn't guess. Questions arrive as cards with options. "Next screen?", "Write this as an annotation?", "Which of these flows matters most?"
• Dev-ready output, not chat. Findings you accept become Dev Mode annotations on the right nodes: what, behaviour, states, tokens, edge cases.
• Your Claude, your bill. Runs on the Claude Agent SDK with your own Claude Code login or API key. Pick model and effort in the plugin. Nothing leaves your machine except the calls to Anthropic.

Quick start

You need **Node 22+**, the **Figma desktop app** (the browser version cannot reach a local process), and **Claude Code signed in** (run `claude` once) or an `ANTHROPIC_API_KEY` in your shell.

**1. Install and start the bridge**, and keep the terminal open:

npm install -g @sesori/figma-review   # once
sesori-figma-review                   # every time you