Every AI request is counted, checked and allowed — or not.
Backapp records each supported workflow request against the user who made it. Free allowances, subscriber limits and hard stops are enforced before the AI provider is ever called — so costs can’t run away.
{
"user_id": "user_8f2a…",
"plan": "free",
"used": 3,
"limit": 5,
"remaining": 2,
"allowed": true
}{
"user_id": "user_8f2a…",
"plan": "premium",
"used": 42,
"limit": 300,
"remaining": 258,
"allowed": true
}Free tier without the risk
Give new users a few AI requests to try. The allowance is enforced per user, so a free tier can’t become an open proxy.
Limits per plan
Set a different allowance for free users and subscribers. The subscription status on the user decides which one applies.
Checked before the provider is called
A request over the limit is declined by Backapp and never reaches the AI provider — no cost, no partial result.
Set the limits. Backapp counts.
- 1Define an allowance per plan.In your project settings: how many requests a free user and a subscriber get, and over what period.
- 2Call workflows as usual.Nothing changes in your app code. Backapp resolves the user, picks the plan from their subscription status and checks the count.
- 3Handle the limit response in your UI.Over the limit, the API returns
429 limit_reachedwith the remaining count and reset time — show your paywall or a “come back tomorrow”.
// CONTENT PENDING: replace with the verified example
const res = await fetch(`${BACKAPP_API_URL}/ai/transcriptions`, {
method: 'POST',
headers: { Authorization: `Bearer ${token}` },
body: form,
});
if (res.status === 429) {
const { error, remaining, resets_at } = await res.json();
// error === 'limit_reached'
showPaywall({ resetsAt: resets_at });
return;
}
const job = await res.json();{
"error": "limit_reached",
"plan": "free",
"limit": 5,
"used": 5,
"remaining": 0,
"period": "month",
"resets_at": "2026-12-01T00:00:00Z"
}What Backapp tracks per user.
Usage is a counter on the user record, kept by Backapp. Your app can read it to show remaining requests; it never writes it.
plan- free or the subscribed plan, from subscription status.
limit- Requests allowed for this plan in the current period.
used- Accepted requests so far this period.
remaining- What your app can show the user.
resets_at- When the counter returns to zero.
Request counts per period
A request allowance per plan, reset daily or monthly. Enough for a free tier, a subscriber limit and a hard cap on cost.
Credits and per-workflow limits
Weighted credits (an image costs more than a transcription) and separate limits per workflow are being scoped and will be listed in the docs when confirmed.
Limits read from the user — not from your app.
The counter is tied to the identity, the plan comes from the subscription, and every AI workflow checks both before running.
Questions about limits?
Is a declined request charged?
No. Limits are checked before the AI provider is called. A declined request costs nothing and returns immediately.
Can I show users how many requests they have left?
Yes. Remaining count and reset time are available on the user record and in every limit response.
What happens when a free user subscribes?
Their subscription status changes, so the subscriber allowance applies from the next request. The counter itself is not reset.
Can I set a limit per workflow, or use credits?
Per-workflow limits and weighted credits are being scoped. Today the allowance is a request count per plan and period.
Can a user bypass the limit by reinstalling?
A reinstall creates a new anonymous identity today. Sign-in flows that carry identity across installs are listed in the docs as they are confirmed.
Ship a free tier without a runaway bill.
Identity, subscriptions and AI workflows from one backend.