Draftbit Review: AI App Builder, Pricing & My Test Results

Draftbit is an AI app builder for mobile apps that also run on the web. You describe the app in a chat, its agent writes the code with React Native and Expo, sets up a Supabase database and shows a live preview, and the same app runs on the web, iOS and Android. You can also edit the screens by hand in a visual editor or a code editor.

The free plan gives 3,000 credits once and publishes web apps to a draftbit.dev address. Standard is $15 a month and adds code export, a custom domain and store publishing; Pro is $30. Paid plans can use your own Claude or ChatGPT subscription for the agent.

Every number here comes from my own tests and my own monitors. I used the free plan only: no card, no paid credits, no trial. Every app builder gets the same brief, checks, changes and attacks. Where the free plan stopped a test, I say so and leave that test out.

Draftbit has a free plan, then $15 Standard a month, or $12 billed yearly, and $30 Pro. Prices read on 27 October 2026, monthly view. The free plan gives 3,000 credits once. Any paid plan can add credits, from $5 at 1,000 credits a dollar.

Draftbit pricing page, monthly view: Free at $0 with 3,000 credits to get started, Standard at $15 a month and Pro at $30 a month, with the features of each plan

Here is the number you will ask about first. In total I paid $0. The free plan gave 3,000 credits, and I used 2,823 of them: about 1,530 for the first build and 200 to 420 for each of four changes. I stopped with 177 credits left, because the next change would have taken me close to zero and the screen says extra charges apply when credits run out. If you want more, you pay.

Draftbit credit usage grouped by thread: the first build 1,534 credits, rename instructor role to coach 421, member phone collection 386, about page 278 and colour scheme 204, a total of 2,823 credits

Draftbit sells credits from $5 at 1,000 credits a dollar, so what I used is worth about $2.80. A build and all twelve changes would need roughly 5,400 credits at the rate I measured. I did not test what 1,000 bookings a month would cost.

The agent built the whole app in 22 minutes from one prompt, and it asked me nothing. It chose the design, wrote the booking rules into the database and added nothing I had not asked for. It also said plainly that payments and emails would need keys from me.

Draftbit test app home page in dark green and cream: Fitting Room, Make room for movement, a green card saying Your next good habit starts here and a 5 classes for $95 card, above the weekly timetable

To be live, I then had to do three things by hand: set the production database address and key, deploy the database to production and, once the app was published, make the first admin in the database panel. The app was live 43 minutes after I submitted the brief.

All of it: 24 of 24 checks I could run passed. Booking, My bookings, cancelling with the credit coming back, the full-class rule, the coach's roster with attendance and the admin's class tools all worked.

Six checks, the ones about paying with Stripe and sending emails, were not run, because they need my own keys. You should test them on your own accounts. Nothing was invented: the three classes, the prices, the address and the phone number match the brief.

Draftbit test app member timetable in dark green and cream: five class rows from Monday to Saturday with times, instructors, places available and Book class buttons, and a View booking button on Power Reformer

The booking window shows its text on top of the page behind it, so the words overlap. It still books, but a real customer would notice.

I ran four of the twelve changes and all four worked first time. They were a dark green and cream colour scheme, a phone number on sign-up with an admin Members page, the role renamed from instructor to coach, and an About page. After I republished, the old bookings and logins still worked. Then the free credits were nearly gone, and this test is not graded.

Draftbit test app admin Members page in dark green and cream: Your studio community, a search box and four members with their emails hidden and a phone number column, one member showing +1 512 555 0123

A change cost between 204 and 421 credits, so a change costs about a tenth of the free allowance. The agent told me each time to deploy the database to production before the change would reach the live app.

Publishing worked on the free plan. Publish put a web app on a draftbit.dev address in three minutes, with no paywall. A custom domain, the simulators and the test builds are the paid parts, and the menu shows a padlock on two of them.

Draftbit Publish Project menu: Progressive Web App published as version 1.1.0 with Publish and Unpublish, Apple App Store, Google Play Store and Over-the-Air Update with Setup buttons, Simulators and Test Builds with padlocks, and a Supabase database Review and deploy button

There was no sideways scroll at 360, 768 or 1440 pixels. The booking window overlaps its own text on a phone too. I measured speed on my own computer, because Google's service refused the request that day, so treat the number as a guide.

Every day I also run the Draftbit speed test and the Draftbit security checks.

A review tells you what happened on one day. This tells you what happens every day after. My own monitor opens the app every five minutes for 30 days, from 27 October to 26 November 2026, and records whether it answered and how fast.

The app sits on Draftbit's own address. If Draftbit removes a free app that nobody pays for, the record will show it. That is a finding too.

You can open the record and see every check.

Every check so far is on the Draftbit uptime record.

Both of my attacks were blocked. A member asking for another member's booking got an empty answer, cancelling it was refused, and a member calling the admin delete-class function was told admin access is required.

Nobody can make themselves staff. A coach or admin sign-up creates a member with a request waiting, and an admin approves it. The first admin has to be set by hand in the database panel. The password reset email arrived within seconds and the new password worked.

Draftbit test app sign-up form with Member, Coach and Admin tabs, the Coach tab chosen, a Which coach are you question with Lena Park and Tomás Reyes, and the note that staff accounts need approval from the studio before they can use coach or admin tools

Draftbit's own scan flagged more than ten warnings when I deployed the database, all about functions that signed-in or signed-out users can run. None of them let my attacks through, but you should read that list on your own app before you go live.

Yes: six of six data checks passed. Bookings, attendance and the Ask the studio form wrote real records. After I republished the app as a new version, with the colour change in it, every booking, credit and attended mark was still there. The admin could view bookings by status, edit and delete classes, and download all bookings as a CSV.

Draftbit Supabase database panel on Production: six tables named profiles, classes, bookings, orders, studio_messages and email_outbox, with row counts, and Query your database and Deploy to production buttons

Deleting a class cancels its bookings and gives the credits back, and the app says so before you confirm. The database is Supabase, which is free during Draftbit's beta, and I could not find what it will cost afterwards.

I did not run payments or emails, so this test is not graded. The app has a pass page, but it says online checkout is waiting for Stripe test setup and no payment can be taken. The agent told me it needs Stripe and email keys, which I did not supply.

Draftbit test app class passes page: the 5-class pass at $95 and a single class at $24 with their Buy buttons switched off and a message that online checkout is awaiting Stripe test setup

Booking emails are saved in a queue and the booking screen says so; they are sent only after an email key is added. The password reset email did arrive, because Supabase sends it.

I did not export the code, so this test is not graded. On the free plan the Download option is locked behind a padlock, and the pricing page lists code export and GitHub export from Standard, at $15 a month.

Draftbit Export dialog with a Select destination list showing Download with a padlock

The code is React Native with Expo, and the database is a Supabase project, which Draftbit says you can claim as your own. I did not test either.

I asked Draftbit support the three questions twice on 27 October 2026. One was an email to its support address. The other was a message in the chat box inside the product, which said the team would look into it and showed Waiting for a teammate. No answer had come back by the time I finished testing, so support is not graded and I will update this when one arrives.

The cancel path is stated on Draftbit's pricing page: a downgrade or a cancel takes effect at the end of the current period, and you keep the paid features until then. I never paid, so I did not test it. The chat menu has an I want to cancel my subscription button.

Draftbit support chat: the three questions I sent, the automatic reply that the team will look into it right away, and the words Waiting for a teammate

Draftbit has a Trustpilot TrustScore of 4.5 from 37 reviews, read on 27 October 2026, with 67 percent giving five stars. That is a small number of reviews. People praise how easy it is to build without code. The few complaints are about being held behind a paywall and about small layout problems such as overlapping items, which matches what I saw in the booking window.

If this were my studio, I would use Draftbit to get a booking app built in one sitting, and I would plan to pay for a month to take the code and add a domain. It built the booking rules, the capacity limit, the three roles and the admin tools from one message, kept staff access behind an approval, and published for nothing.

It suits people who want a working mobile and web app quickly and accept paying for the export. It suits you less if you want to build and edit on the free plan alone, because 3,000 credits covered a build and four changes. It also suits you less if the page must load fast on a phone, or if you cannot spend an hour wiring the keys, the database and the first admin. I could not test payments, emails, the other eight changes or the code export.

The six graded tests average 4.2 out of 5, and I rate Draftbit 4.2. Accuracy, security and the database are the highest, because one prompt made a working app that held up against both attacks and kept its data. Publishing and cost are the lowest. On a paid plan the changes and the export tests could be run, so treat this as a free-plan rating, from $15 a month.

See the Draftbit test results and how I rate.