Every screen, side by side
The whole product on one infinite canvas. Live cursors show who is looking — people and agents.
Free plan, no card. Unlimited MCP calls on every plan.
Alone, an agent guesses: a template, placeholder numbers, a screen that could belong to any bank. Inside Croqui it runs with your skills — brief, components, design system — and writes this product's screen.
One command in Claude Code. Codex, Cursor and any MCP client work too — see the docs.
“Design the checkout on Croqui.” The agent reads your brief and components, then builds it region by region while you watch.
Click the screen to leave a note. The agent resolves it and replies. Approve the version you want, and that is the one that ships.
Croqui has no model of its own and resells no tokens. It is one MCP server: any agent that speaks it draws on the canvas, reads your notes and writes the files.
Any MCP-capable client works. Add the Croqui server once, and the agent gets the canvas, the screens and your notes as tools.
Your agent writes files. You look, point and approve. Nobody re-implements a mockup.
The whole product on one infinite canvas. Live cursors show who is looking — people and agents.
Each frame is a .tsx file. Desktop and mobile in one project, ready for your repo.
The agent reuses your buttons, inputs and prices instead of inventing new ones every prompt.
Drop a pin where it’s wrong. The agent fixes that spot and replies on the note.
Watch the screen change as the agent writes it. A badge marks the frame it is touching.
Publish the bundle and run the same approved screens in another app.
Every write is versioned. Roll back a bad edit without leaving the canvas.
Your agent writes the backend fine. The UI comes out looking like every other AI app. Croqui gives it your components, and gives you a way to say “this, here”.
Free or Pro →Designer, PM and engineer review the same live screens the agent is writing. Whoever has approval access signs off, and the approved screen goes straight to engineering.
Team →Clients leave notes on real screens as free viewers. No more feedback on screenshots in email threads, and no mockup to rebuild afterwards.
Team →No token resale, no per-call quota: you bring the agent, Croqui is the canvas. Yearly is two months free.
Try it on one screen with your agent.
For shipping a whole product with your agent.
For teams and agencies reviewing with clients.
Prices in USD, billed via Stripe. Cancel anytime. Need SSO, an audit log or 15+ seats? Ask about Studio.
Croqui.dev is a design canvas for the coding agent you already use.
A design canvas where your coding agent writes real React screens that render live. You pin notes on the screen, the agent resolves them, and you approve the version that ships. Every frame is a .tsx file that reuses your components, so nothing gets re-implemented from a mockup.
No. Croqui.dev is a canvas, not another model. You connect the agents you already run — Claude Code, Codex, Cursor, Grok or any MCP client — over a single MCP server.
In Claude Code, install the Croqui.dev plugin, which brings the MCP server plus the agent skills: claude plugin marketplace add croqui-dev/croqui-mcp, then claude plugin install croqui@croqui. For the MCP server only: claude mcp add --transport http croqui https://croqui.dev/mcp. In Codex: codex mcp add croqui --url https://croqui.dev/mcp. Full steps in the docs.
As many as you want. MCP calls are unlimited on every plan, including Free. There is no per-call quota and no token resale.
No. Free and Pro are one seat: you and your agent. Team adds a shared canvas, free viewer seats so clients can leave notes, and automatic delivery of approved screens to engineering.
Figma produces a picture that someone re-implements later. A prompt-to-app builder uses its model and its component defaults. Croqui.dev uses your agent and your design system, and puts a human approval gate between the screen and production.
Yes. Point your agent at a screenshot, a live page or your components and ask it to import them. It inventories what it sees, you pick what to bring in, and every screen after that reuses it.
Not to review a screen — the canvas renders it live, and you annotate it with pins. The files the agents write are React, so an engineer can take them as they are.
More in about Croqui.dev and the docs. Machine-readable summary: llms.txt.