It plans the work first.
A build plan lists each step and ticks it off as it lands, so you can follow along and stop at any point.

Sanic AI
Tell Sanic AI what you want. It plans the work, builds it on the same canvas you edit, and checks its own result. You keep every lever: the model, the effort, the undo.


What it does
One request can carry a whole app, or one small change to the part you have selected.
A build plan lists each step and ticks it off as it lands, so you can follow along and stop at any point.

Screens arrive on the canvas as real components. Logic arrives as TypeScript you can read, including code for your connected Supabase tables.

Every change is validated and compiled before it lands. An optional design review looks over the finished screens.

Build plan and Design review are optional settings in Sanic AI.
Models and effort
Three models, each with three effort levels. Smaller requests cost less; the deepest work gets the deepest model.
| Model | Best for | Included from |
|---|---|---|
| Lite | Small, economical changes: a label, a colour, one component. | Pro |
| Core | Everyday building: new screens, flows and services. The default. | Pro |
| Pro | Complex changes that touch many screens, code and data at once. | Premium |
Effort
A fast pass for small requests.
The everyday setting.
More reasoning for the hard ones.
Models by plan
You stay in control
Sanic AI changes your app the same way you do, through one checked door.
A change that breaks the rules is refused whole. Nothing lands half done.
An applied change joins the shared history with your visual and code edits, so one undo takes it back.
Keep editing while it works. If your edits overlap its proposal, you choose what happens; nothing is merged in silence.
Sanic AI sees that a service is connected and what it can do, never the keys. If a request needs a new service, it asks you to connect it.
Your current page and selection travel with each request. It reads project files and component definitions, not your users' data.
Stop cancels the running request. Separate conversations keep separate threads of work on the same project.
Credits, plainly
Credits track what Sanic AI actually reads and writes, at the chosen model's rate. Visual editing and the code you write yourself never spend one.
From 200 to 1,500 a month, depending on the plan. Included credits reset each billing period.
Optional packs start at 150 credits for US$10. Purchased credits keep until you use them.
A free account gives you the editor, the code workspace, themes and connectors today.