Usage allowance: a daily and a weekly limit, measured by what builds cost
Anticode doesn't count builds or sell credits by the prompt. Every model build spends part of a daily and a weekly allowance according to what it actually cost to run, in tokens at the model's price. A small change uses a sliver; a large app uses more. Builds that are refused or fail cost nothing and say Not charged. Amounts for each plan are on the pricing page.
Why measure by cost instead of counting builds
A flat "50 builds a month" sounds simple, but builds aren't alike. Changing a headline and writing a whole browser game from scratch differ in cost by a large multiple, so a per-build count either overcharges small edits or can't cover large apps. Measuring by what each build used is fairer in both directions. Anticode's AI website builder usage limits work the same way: a share of an allowance, with the reset time shown.
Two windows
| Window | Resets | What it's for |
|---|---|---|
| Daily | Every day at 00:00 UTC | Spreads use across the week, so one long session can't use it all at once |
| Weekly | Every Monday at 00:00 UTC | The overall amount of building in a week |
The builder shows each window as a percentage used, with the time it resets. The headline meter shows whichever of the two is higher, since that's the one that would stop you first. When either reaches its limit, model builds wait for the reset, or are paid from credit packs if you have any.
Before, during and after a build
- A quote first. Before each build, the builder shows roughly how many tokens it will use and what share of your daily limit that is. An app build is quoted higher, because the model writes the whole program. It's an estimate; the build is charged for what it really uses.
- A reservation while it runs. A build in progress holds back an estimate, so two builds started at the same moment can't both slip under the limit. If a build is still unfinished after fifteen minutes, it's marked as failed and stops holding anything.
- A charge only on success. When the build finishes successfully, its actual cost is counted. A refused or failed build counts as nothing, even if the model used tokens.
- A receipt either way. Every build gets a receipt with the model, the tokens, the cost and its share of each limit; failures show Not charged. Receipts are covered on build receipts.
If several builds fail in a row, model builds pause for a few minutes. That's there to stop a repeating failure from eating into credits you've bought.
How the plans relate
- Free
- A small, time-limited allowance for trying model builds. Once it ends, model builds need a paid plan or credit packs; preview builds stay free.
- Pro
- The standard allowance for one person building regularly.
- Max
- Pro with four times the allowance, for people who build a lot.
- Team and Agency
- A shared workspace with a larger allowance. Anyone working in it builds at the more generous of their own plan and the workspace's.
Past the allowance, credit packs take over. They're bought once, never expire and are spent only when the plan's allowance for the day or week is used up, so they never replace what your plan already includes.
What doesn't use the allowance
- Preview builds from the design engine, which run in your browser without an account or a model.
- Templates. Using one copies it into your project with no model call.
- Your own edits in the visual editor, layout editor and code editor.
- CLI pushes of your own changes to a project's spec.
Two things do count like a build: questions to the analyst on the Intelligence page, and model-written apps and changes started through the MCP server, which quote first just as the builder does.
Limits
- Estimates can be off. The quote is a guide based on typical builds. A long request or a large app can use more.
- UTC resets. The reset times follow UTC, not your local midnight.
- A confirmed email is needed before model builds can start.
- No amounts here. Allowances and prices change with the plans, so this page doesn't quote them; pricing has the current figures.
Questions people ask
Do failed builds count against my limit?
No. Only successful builds are counted. A refused or failed build shows Not charged on its receipt and takes 0% of either limit.
When does my allowance reset?
The daily window resets at 00:00 UTC every day and the weekly one at 00:00 UTC on Mondays. The builder shows both reset times.
Why did an app build use more than a page build?
An app is a whole program the model writes, so it uses many more tokens than planning a page. The quote shows the difference before you start.
What happens when I reach the limit?
Model builds wait until the window resets, unless you have credit packs, which are spent only once the allowance is used up.