AI writes more than the pages. The backend code too.
Sign-ups, saved enquiries and bookings, taking money — normally someone has to write a backend for that. Here you describe what it should do, and AI generates the server code, the data tables, and the login into the same project. You don't need to know how to code.
Direct answer
Can I build a site with login, forms, and payments on Creght without writing code?
Yes. The backend code is generated by AI — you say what you need in the chat, for example "add a booking feature and save what the customer fills in", and AI writes the server code, creates the data table, and wires the pages up. You never have to read that code. Four common recipes are also built in: create login, save form submissions to a table, generate CRUD handlers for a table, and require login before a page is viewable — one click generates them against your current project. Login supports email and password plus Google and GitHub OAuth, Alipay web payments have a full guide, and third-party apps such as Resend can be connected for email.
The generated code stays visible and editable. Func is the project's backend code entry point, holding backend logic and API handlers; it reads and writes the data tables and the user system. Edit it directly in the editor if you want to, or pull the project down with the Creght CLI and keep going in Claude Code, Cursor, and similar environments.
The limits are stated too: data tables are JSON tables and do not support joins, aggregation, transactions, or fuzzy matching — systems that need those should use a real database called from Func. The user directory is scoped to the whole project, so any who-can-query-whom rule has to be spelled out in code (AI can write that part as well). Func has a daily call limit that varies by plan.
The problem
What gets stuck when you can't write code
A marketing site eventually needs login, bookings, leads, or memberships. All of them need backend code — and the traditional answer is a second project for someone else to write and maintain.
| Without Creght | With Creght |
|---|---|
| Can't code, so the backend gets outsourced and every change waits on someone. | You describe what it should do; AI writes the backend code, and the changes too. |
| Need login? Stand up a service, model users, write sessions. | Tell AI you want sign-up and login; it generates the server code and the pages. |
| Form submissions only send email, and data lives in an inbox. | AI writes the handler; submissions land in a data table you can query and edit. |
| Members-only pages mean writing your own middleware. | Click "Protect page" in the panel and AI generates the login requirement. |
| Taking payment means a gateway, callbacks, and signature checks. | Alipay web payments have a full guide, and the integration code goes to AI too. |
How it works
You say what it should do, AI writes the backend
AI writes the backend code — you never have to read it
Tell AI what the site should do: who can log in, what information to keep, what happens after a submission. It writes the server code, creates the data tables, and connects the pages. The code shown here is what AI generated for a booking feature. It sits in your project — ignore it entirely, open and edit it whenever you like, or pull it down with the Creght CLI and keep developing.
/backend/func/booking.ts
AI-generatedexport function create(input, ctx) {
const user = ctx.auth.requireUser()
return ctx.db.insert('appointments', {
startAt: input.startAt,
userId: user.id,
})
}Called from a page
await invoke('booking.create', input)This code was generated by AI and runs straight from your project. If you do read code: platform capabilities come from ctx — ctx.auth for the current user, ctx.db for tables.
The four most common jobs are one click away
Create login, save form submissions to a table, generate CRUD handlers for a table, and require login before a page is viewable. Each is a single entry in the backend panel: one click and AI generates the matching code and page changes against your current project, with nothing to describe from scratch.
Common recipes
Visitors can register and sign in to your site
Site users can register and sign in with email and password, and Google and GitHub login are supported. Who signed up and when is listed in the users view of the backend panel. One thing to note: the user directory is scoped to the whole project, so any who-can-query-whom rule has to be spelled out in code — a step AI can write as well.
Welcome back
Sign in to your site account
Email address
Password
No account?Sign up free
Users
The user directory is project-wide, so any who-can-query-whom rule has to be spelled out in code — a step you can hand to AI too.
Email, payments, and third-party services connect once
Form submissions can be stored; email can go through third-party apps such as Resend, with the key entered once in the panel; Alipay web payments have a full guide and other channels follow the same pattern. One boundary worth stating: data tables are JSON tables — no joins, aggregation, transactions, or fuzzy matching. Systems that need those should use a real database called from the backend code.
Name
Unique per project — tells two configs of the same app apart.
Channel tag
API key
Sender
Related docs
Step-by-step guides
Build the site first, then give it a backend.
Describe what you need and let AI build it; when you want login, form storage, or payments, just tell it next.