rocket.new test data

A Stripe test purchase, a declined card and the two emails, which need your own keys.

Not run. The agent built a pass purchase that says Payment is simulated and Stripe is not connected, and the booking emails are saved in an in-app outbox, not sent. Checks C24 to C29 are left out and this test is not graded. The password reset email is covered under the checks.

• Left out like on every builder: payments and emails need the owner's own keys.

The twelve scripted changes, one prompt each, on the live app.

Not run. The two prompts used all 20 free credits. When I sent the first change, the colour scheme, rocket.new answered that I had exhausted the credit limit and asked me to upgrade. I clicked nothing.

• The credits are the limit of the free plan, not a test fault.

Download the code, unpack it in a clean folder, run it against a database of mine and list what still needs the vendor.

Not run. The download button in the code view opens a window that says Own your code: Download, Deploy anywhere, Modify however you want, with a button that says Upgrade my plan. The files themselves can be read in the code view: a Next.js 15 and React 19 project with Tailwind. I asked support how to export the code and the data.

• The terms were not read.

One real question by email to support@rocket.new, the address on rocket.new's help pages, sent at 21:16 UTC on 27 October 2026: how to stop people choosing the admin role at sign-up, how to export the code and the data, and how to cancel.

No person had answered when this review was written, so support is not graded. The only message from rocket.new was an automatic notice at 21:57 UTC about the failed Update, which asked me to reply if I wanted it investigated. The free plan has nothing to cancel: Settings, Subscription shows the plan, the credits and an Upgrade button.

• Graded when a reply arrives.

The Fitting Room brief was pasted word for word (1,689 characters) into rocket.new's home screen on 27 October 2026, in Build mode, with no answers to give because rocket.new asked no questions. The clock stops when the app is live and the checks pass: 25 of 30 on the full scale, applied here in proportion to the 24 checks that could run.

rocket.new rated the prompt 100 percent, wrote a plan and built the first draft in about 7 minutes, for 7.35 of my 20 free credits. The draft was a front end with made-up data: it said plainly that sign-up, log-in, password reset, the instructor and admin screens, the CSV, saved inquiries and the emails were not built, and that they needed a follow-up request.

I sent that follow-up, listing those parts again from the brief, about 10 minutes later. It took about 8 minutes and the last 12.65 credits, and it built every part inside the app, with no outside accounts. The app was then live on rocket.new's own free address, about 20 minutes after the brief, and 22 of the 24 checks passed.

• My second prompt was needed because the first draft left out half the brief; it is counted as a prompt. The app keeps its data in the browser, which is graded under the database, not here.

The 30 written checks, run by hand against the live address on 27 October 2026. The six payment and email checks (C24 to C29) need your own Stripe and email keys, and the app only simulates them, so they were not run. That leaves 24 checks, graded in proportion: 5 needs 93 percent, 4 needs 80 percent.

22 of the 24 passed. Sign-up, log-out and log-in worked for a member, both instructors and the admin; the timetable showed the three classes, times, lengths and places left; booking, My bookings, cancelling with the credit coming back, the full-class rule (a seventh member on the six-place Power Reformer saw Class full) and the no-credit message worked; the instructor saw only their own classes and the members booked on them, and Mark attended was saved; the admin added, edited and deleted a class, saw every booking and a count per class, downloaded a CSV with eleven columns, and saw a visitor's Ask the studio message.

Two checks failed. The password reset sends no email: the link is shown on the screen instead (C03). And the app invented things the brief did not give (C30): a description and a level for every class, such as Intermediate for Power Reformer, and a made-up member, Priya Raman, with credits and bookings already in place. The class names, times, prices, address, phone and email match the brief.

• All 22 passes were run in one browser. What that means for a real studio is under the database.

Sign-up, log-in and password reset for each role; then the two attacks from the brief, adapted to an app with no server: a direct visit to the admin page as a member, an edit of the browser's stored copy of my own role, a sign-up that picks the admin role, and a second browser looking for the first browser's data.

Roles are cosmetic. The sign-up form lets any visitor choose Member, Instructor or Admin, and an account made as Admin landed on the admin screen at once. A member who typed the admin address was stopped with a page that says it is not available, but the same member changed one value in the browser's stored data, reloaded, and had the admin screen. The password reset shows the reset link on the screen to whoever asks for any email address.

The agent had said so: the sign-up lets anyone pick the instructor or admin role so every view can be tried, and a live studio would restrict that. The login page also prints the four demo accounts and the shared password.

• rocket.new's own security scan was not run. I did not capture network requests, because the agent says the app has no server.

The six data checks: the form and bookings write real records; they survive a reload, a redeploy and a design change; the admin can view, filter, edit and delete them; they export as a file. I added a test in a second, clean browser to see whether another person would see the same records.

Four of the six could be run, and two passed. Reloads kept everything and the admin's CSV downloaded with the right columns. But the records are not real in the sense the brief means: the agent says the app has no shared server, and in a second clean browser the admin saw only the two made-up bookings, not the eight members I had created, and none of the Ask the studio messages. The admin can view, edit and delete classes, but cannot filter the bookings list.

Not run: the redeploy check, because the second Update failed with a compile error, and the design-change check, because the free credits were gone.

• The builder's own data panel was not opened; rocket.new does not offer one for this app, as far as I could see.

Launch was used on the free plan with the Rocket hosted domain. Speed was Lighthouse mobile, three runs on the home page; layout was checked for sideways scrolling at 360, 768 and 1440 pixels.

Publishing worked on the free plan, with no card: the Staging tab put the app on a fittingroom address under builtwithrocket.new in about a minute. The Production tab asks for a custom domain of my own, or to buy one, and I did not use either. The address opens for anyone, and every page I opened carries a small Built with rocket.new badge in the corner.

The three Lighthouse runs scored 92, 91 and 91, so the median is 91. The main content took about 2.9 to 3.1 seconds on a simulated phone, the page weighs 0.36 MB, and layout shift was close to zero. There was no sideways scroll at 360, 768 or 1440 pixels. A second Update, after the credits were gone, failed with Could not launch your app and a compile error about a font, and rocket.new then emailed me that it had found a publishing issue and would look into it if I replied. The first live version kept serving.

• No custom domain, no real phone, no 100 seeded classes. The address is protected by a security check that stops scripts, so my uptime monitor may not be able to read it.

The pricing page, the account page and the credit meter were read on 27 October 2026 on the free plan, with the credits read after each prompt.

I paid nothing and entered no card. The free plan holds 20 credits that never expire, and I used all 20 in two prompts: 7.35 for the first draft and 12.65 for the second. After that rocket.new asked me to upgrade.

The paid plans are a monthly platform fee with credits on top: Pro is $25 for 100 credits a month, Rocket is $50 for 250 and Booster is $250 for 1,500. The page also says you can buy more credits at any time, and I saw no spending cap. I did not measure what a change costs, so I cannot say how far 100 credits go for the twelve changes, or what 1,000 bookings a month would cost.

• No paid plan was bought and no card was entered. Yearly prices were not read.