Rate limits

UnifAPI applies one request limit: 60 requests per 60 seconds, per API key. It is scoped to the key, not to the workspace or the endpoint — two keys in the same workspace do not compete for the same budget, and no operation has a tighter limit of its own.

Headers

Every response — 2xx, 4xx, or 5xx — carries the quota that applies to it, so you never need a probe request to learn where you stand.

A response that never resolved an API key — a 401, for example — reports the default policy with a full budget. It describes the quota you will face once authenticated, not usage you have spent.

Browser callers can read all of these: they are listed in Access-Control-Expose-Headers on cross-origin responses.

HTTP/1.1 200 OK
RateLimit-Policy: "api-key";q=60;w=60
RateLimit: "api-key";r=57;t=60
RateLimit-Limit: 60
RateLimit-Remaining: 57
RateLimit-Reset: 60

The window

The window is anchored on the key's most recent request: the counter clears once a full window passes with no calls from that key. There is no burst allowance, so design for the steady-state rate rather than for a saved-up balance.

Handling 429

async function call(url: string, init: RequestInit, attempt = 0): Promise<Response> {
  const res = await fetch(url, init);
  if (res.status !== 429 || attempt >= 5) return res;

  const wait = Number(res.headers.get("Retry-After") ?? res.headers.get("RateLimit-Reset") ?? 1);
  await new Promise((r) => setTimeout(r, wait * 1000));
  return call(url, init, attempt + 1);
}

Two rules to live by:

1. Pace before you are throttled. Read RateLimit-Remaining on every response and slow down as it approaches 0. Waiting for a 429 to start pacing wastes a round trip on every worker at once.
2. Honor Retry-After. It is the wait the limiter actually needs; guessing usually makes things worse. If it is absent, fall back to RateLimit-Reset. Raising concurrency after a 429 never helps — the limit is per key.

Asking for more

Limits are set per key, so a higher limit is a configuration change rather than a plan upgrade. Email support@unifapi.com with your workspace ID, the Skill or operation you are running, and a rough QPS estimate.