Rate limits & quotas

Two independent controls apply to every API key:

• Rate limit — sustained requests per second, enforced as a token bucket with burst capacity equal to your plan's rate. Short bursts up to the limit are fine; sustained excess is rejected.
• Monthly quota — total successful requests per calendar month, shared across all APIs. Quotas reset on the first of the month at 00:00 UTC.

Plan limits

This table is generated from the same configuration the gateway enforces:

Response headers

Every rate-limited response — success or rejection — carries the current state:

x-ratelimit-limit: 15          ← plan's sustained requests/second
x-ratelimit-remaining: 14      ← tokens left in the bucket right now
x-ratelimit-reset: 1785503186  ← epoch seconds when the bucket refills
x-quota-limit: 250000          ← monthly quota
x-quota-remaining: 249862      ← requests left this month
x-quota-reset: 1785542400      ← epoch seconds at the next monthly reset

When you exceed a limit

There are no automatic overage charges — when a quota is exhausted, requests are simply rejected until you upgrade or the month rolls over.

Client best practices

• Watch x-ratelimit-remaining and pace proactively instead of reacting to 429s.
• On 429, sleep exactly Retry-After seconds — it is calculated, not a guess.
• Add jitter when many workers share one key, so refilled tokens are not consumed in a stampede.
• Alert your team when x-quota-remaining drops below ~10% — before production traffic stalls.