Staged agent
You are not waiting for a monolith dump. Screens arrive in order.
Use case
Figma cannot tell you the form validation is wrong. A staged Cursor build can. Get the spine up, click it, argue, restore, send the next prompt. Hours, not a sprint of “we’ll know after we build it.”
Stages
Shell before everything
Stop
Kill a bad run
Restore
Checkpoints
Same project
No fork soup
Why this exists
The point is the click path. Build home, the core object, the failure case. If the idea dies, you spent tokens, not a quarter. If it lives, you already have TypeScript.
Spine
In the product
You are not waiting for a monolith dump. Screens arrive in order.
When it is clearly wrong, stop. Do not let it finish a mess.
Label before a wild idea. Restore after.
@path so the next turn sees reality.
Edit a user message and resend from a checkpoint if the studio supports your flow.
Iframe on a call. No “imagine if this button worked.”
How
01
What must be true for this idea to be worth a team.
02
Cut the settings page. Build the bet.
03
Watch the first thirty seconds.
04
Restore and try the other bet if it died.
Questions
Real Vite + React + TypeScript files in a project you own. Preview is a running Vite server proxied through the studio iframe.
The agent stops until the month resets or you upgrade. Pro is $19.99/month via Whop card checkout. The webhook sets your plan to Pro.
Yes. Export a zip from the studio. Secrets are not stuffed into the zip as plaintext from the ciphertext store — add env files yourself after export.
Rough copy is OK. Broken back buttons are not. The prototype has to click.
Export, or keep iterating until it is the product. Founders often do the latter longer than they admit.
Next
Free is live. Pro is $19.99/month on Whop. The studio, preview, and export are on this host — not a waitlist.