Skip to main content
Quotas prevent one runaway script or misbehaving key from draining your plan. They cascade from broad to narrow:
The tightest value wins. If your plan allows 120 req/min but a key preset says 30 — that key is capped at 30.

​
What you can limit

​
Presets

Three preset cards, each with link, edit and delete icons on the right. Production reads Live traffic. Generous rate, capped spend, no scheduling from a key, with four chips: Models: All, Schedules: Denied, Budget: 4,000 / monthly, and Rate limit: 120. Prototype reads For experiments — a small daily budget and the cheap models only, with three chips: Models: 3 selected, Budget: 150 / daily, and Rate limit: 20. Scheduled jobs reads May schedule, but no more often than every fifteen minutes, with three chips: Models: All, Budget: 800 / monthly, and Rate limit: 60.Three preset cards, each with link, edit and delete icons on the right. Production reads Live traffic. Generous rate, capped spend, no scheduling from a key, with four chips: Models: All, Schedules: Denied, Budget: 4,000 / monthly, and Rate limit: 120. Prototype reads For experiments — a small daily budget and the cheap models only, with three chips: Models: 3 selected, Budget: 150 / daily, and Rate limit: 20. Scheduled jobs reads May schedule, but no more often than every fifteen minutes, with three chips: Models: All, Budget: 800 / monthly, and Rate limit: 60.

A preset, and what it caps

The chips on a preset card summarise it; they are not a complete inventory. Models, budget and rate limit each show their value whenever the preset sets one, so a missing chip among those three does mean the preset says nothing and the plan decides. Scheduling is deliberately different: a card shows Schedules: Denied only when the preset forbids scheduling outright, because a chip on every card would bury the one card that actually blocks something. And a preset can hold a limit that has no chip at all — Scheduled jobs above permits scheduling but no more often than every fifteen minutes, which is why the card says it in words. Read the description as well as the chips. A preset is a named bundle of limits. Define once in Settings → Quotas and assign to members or API keys.
A New preset dialog with a close X in the top right. A required Name field with placeholder e.g. Developer Tier, and an optional Description field. An Allowed models section with an All models checkbox, checked. A Budget limit (credits) field, empty, beside a Budget period dropdown set to Monthly. A Rate limit (req/min) field, empty. A Scheduled workflows dropdown set to Inherit, with the note This preset says nothing — the plan, and any preset applied before it, decide. A Create button at the bottom, shown disabled because the required Name field is empty.A New preset dialog with a close X in the top right. A required Name field with placeholder e.g. Developer Tier, and an optional Description field. An Allowed models section with an All models checkbox, checked. A Budget limit (credits) field, empty, beside a Budget period dropdown set to Monthly. A Rate limit (req/min) field, empty. A Scheduled workflows dropdown set to Inherit, with the note This preset says nothing — the plan, and any preset applied before it, decide. A Create button at the bottom, shown disabled because the required Name field is empty.

Creating a preset: the four things it can decide

The New preset form decides four things: an absolute credit budget over a chosen period, a requests-per-minute ceiling, which models are allowed, and whether this preset may schedule workflows at all. Common presets to start with:
  • Production — no rate limit, allowed models = whitelist of blessed slugs, high storage
  • Development — 30 rpm, 1M tokens/day, all models allowed
  • Read-only — 10 rpm, no image/video generation, minimal storage

​
Assigning a preset

To a member: Settings → Members → member row → Quota preset. Their API keys inherit automatically unless the key has its own preset. To an API key: API Keys → key row → Preset. Overrides the member’s preset for that specific key.

​
Seeing effective quota

Each key row in API Keys shows the resolved quota — the final values after cascade — with a source column (platform / plan / member / key) so you know where each limit comes from.

​
API visibility

Clients see quotas in error responses:
That arrives as a 403, not a 429, and carries no Retry-After — see Rate limits. And storage-related 413 responses from /v1/files: