Put an AI-generated app online, with real accounts and a real database.
Your evening projects, built prompt by prompt, work in the demo. Here the AI builds them on a foundation that already provides accounts, the database and hosting, and the Workshop shows you what it wired behind each button.
Updated on
Demo: Brasserie du Mont, a fictional craft brewery in Annecy, and its trade portal built with Kubbler.
The home page, in sections
Home hero, Home bento, Home process, Home faq: the page tree lines up the brewery's home page, and each section unfolds down to its blocks.
The page, fixed in place
The 20-litre keg and the case of 12 x 33 cl, as the bars read them. A line of copy or a photo changes right on the canvas.
Behind the order form
The proCatalog block calls three routes, listCatalog, listRounds and placeOrder, which read from and write to the beers, proAccounts and orders collections.
The public site, no trade prices
Bières d'Annecy, delivered on Thursdays: the home page announces the next delivery round and points to the trade portal. No trade price leaves the server for a visitor.
Brasserie du Mont, a demo built with Kubbler.
Brasserie du Mont brews six beers in Annecy and delivers every Thursday to the bars, restaurants and wine shops around the lake. Its app is a trade portal: the brewery approves each account, the bar orders at trade prices excluding VAT, and the Thursday round is drawn from the orders received.
6 beers, by the 20 l keg (€84 to €98 excl. VAT) or the case of 12 x 33 cl (€24 to €29 excl. VAT)
20 approved trade accounts, 3 waiting for their SIRET number to be checked
Orders close Wednesday at noon; the bar pays online or gets a 30-day invoice
Neither the brewery nor the bars exist, nor do their orders; card payments are Stripe test-mode runs, with no real money involved.
The whole site, page by page
Each thumbnail is a complete page of the demo, as a visitor sees it. On a computer, hover one to scroll through it.
A bar's order, from the catalogue to the van.
What a bar does in the brewery's trade portal, and what the AI built for each step, as you read it in Kubbler.
A bar looks through the six beers
Strength, the peak each beer is named after, keg or case. Where the price would be, the page reads "Prix pro sur compte validé": trade price once the account is approved.
You review what the AI put together
The Workshop lays out the home page one section at a time. You rework a line or a photo on the page; trade prices stay in the database, and the server refuses them to any account not yet approved.
Le Comptoir du Thiou fills in its order
Approved account, prices excl. VAT on screen: 2 kegs of Blonde du Semnoz, a keg and 2 cases of IPA des Aravis, a case of Saison, €421.20 incl. VAT for the 22 October round.
You follow "Commander" down to orders
In Mechanisms, proCatalog leads to placeOrder, which records each order as a single document. Two bars confirming in the same second get two separate orders, and neither is lost.
It picks the 30-day invoice
Order BDM-T879X is recorded: it goes out with the Thursday 29 October round, invoice due by 28 November. The other option, a card, is settled on the spot via Stripe.
Each order becomes a row
In the orders collection: reference, delivery Thursday, kegs, cases, totals excl. and incl. VAT, payment method and status. In production, the Workshop shows it without letting anyone edit it.
Screenshots: Brasserie du Mont's trade portal as seen by Le Comptoir du Thiou, a fictional bar in Annecy, then the project open in the Workshop.
Wednesday, 6 pm, Thursday's round is ready: 31 kegs, 54 cases.
Orders closed at noon. The "Tournée du jeudi" screen adds up the week's orders: 18 stops, €4,173 excl. VAT to deliver, 10 orders paid by card and 8 to invoice at 30 days. Nobody copied anything out of a text message: it all comes from orders the bars placed themselves.
Thursday 22 October
At the top, the delivery date, the Wednesday noon cut-off and four figures: kegs, cases, stops, amount excl. VAT.
Paid or to invoice
10 orders settled by card, 8 to invoice at 30 days: the brewery knows which invoices to send after the round.
The load, beer by beer
Blonde du Semnoz, 12 kegs and 10 cases; the "Dans le camion" line totals 31 and 54, what the driver puts in the van.
What your app finds when it arrives, and what you build on top.
The part that kept giving way in your generated projects (sign-in, the database, going live) is written once and shared by every project. The AI works above it, on the part that is actually your idea.
Our part
In every project from the day it is created: the AI relies on it without regenerating it.
Your users' sign-in
Two-factor authentication, a verification email at sign-up, forgotten password: the same code across every project, maintained by our team, instead of one more generated version.
A real database
A database of your project's own, backed up automatically, with a Dev environment for your experiments and a Prod those experiments never touch.
Going live
Hosting with OVHcloud in France, HTTPS, your domain connected to the project, and a cookie banner where everything stays refused until the visitor chooses. No server to rent, no deployment to configure.
Your part
What you decide, by prompt or by hand, and read back before you publish.
The plan before the code
Your description becomes a plan: the pages, the collections and their fields, the API routes, and what the foundation already provides. You correct it before building starts.
Who sees what
A route open to everyone or limited to signed-in accounts, a required field, a default value, pro prices kept for approved accounts: you find each one in the Workshop, and change it yourself or ask the AI to.
Code, when you need it
A custom route or a scheduled task is written by hand, next to what the AI produced, and every save is checked before it is accepted.
Last night your app worked in the demo. It breaks at the first real sign-up.
You know the evening: an idea, a few prompts in Lovable, Bolt or Cursor, and a directory or a small tool running before midnight. The next day a friend signs up and everything slips: sign-in throws an error, the database refuses to write, deployment asks for keys you can no longer find. And when you ask the AI to fix it, it rewrites three files you have never read.
The usual ways out all cost something. Taking over the code yourself means understanding what you did not write. Paying a developer to straighten it out costs more than the project brings in. Wiring one service for sign-in, another for the database and a third for hosting leaves you with three accounts, three bills and three sets of settings to keep in line. Plenty of evening projects stop there.
Kubbler starts from what broke. Sign-in, the database, hosting at OVHcloud and the cookie banner exist before your first prompt, in a foundation shared by every project. The AI builds your pages, collections and routes on top, and the Workshop lets you follow each button to the collection it writes to, before publishing and after.
What it costs
One project and 500 sparks a month, the budget that pays for AI generations. Packs can be added if you need more.
With your own domain: €44 excl. VAT a month.
No commitment. The domain name itself is paid separately.
See all pricesWhat to have ready
The idea, and the people who will use it: who signs up, who searches, who moderates.
What each listing must contain, field by field.
The restricted pages: what needs an account, what is for administrators only.
Screens you already generated elsewhere, as an HTML export or a .zip, if you want to keep their layout.
A domain name, if the project needs its own address.
Nothing else: no specification, no mock-up.
The questions we get asked.
Including the ones where the answer is "not yet".
In Kubbler the app is built straight onto a foundation that is already hosted: you approve the plan, the AI builds, you read it back in the Workshop, then you publish over HTTPS on a kubbler.app address or your own domain. No repository to connect, no server to prepare. Code generated elsewhere, on the other hand, cannot be imported.
Those tools write the code for the whole application, sign-in and data access included, and hand it over to you. With Kubbler, the AI writes only your project's layer, on a foundation it does not touch, and we host the whole thing. The trade-off: that foundation stays ours and the code does not export; your data, though, stays yours.
Its screens, yes: an HTML export or a .zip becomes editable sections, and you are told which parts the conversion could not carry over. Its logic, no: you describe the project in the Studio, the AI rebuilds the collections and routes on the foundation, and you read them before publishing.
Count €29 excl. VAT per month for the Starter plan, with 500 sparks a month to pay for AI generations (an estimate shows before each run). To put it online, hosting costs €5 excl. VAT per month on a kubbler.app address, or €15 excl. VAT per month on your own domain name. Nothing is charged if the AI service fails, and the plan carries no commitment.
No, not today: the foundation stays with us, and no Git repository comes with the project. Your data belongs to you, and your API description exports at any time to OpenAPI 3.0, Swagger or Postman.
Not yet. A Kubbler app takes one-off card payments (Payment brick, through Stripe), with no subscriptions, and sends no text messages or campaigns. It knows two roles, administrator and user. If your idea rests on a paid subscription, better to know before the first prompt.
Pick up last night's idea, this time with accounts and a database.
In the Studio, the AI proposes a plan from your description; you correct it, then it builds on the foundation. Before you publish, the Workshop shows what each button triggers, down to the collection.