Skip to content

Rate limits

The per-key limit, its headers, and the four ways to be told to slow down.

The limit

Calls are counted per API key, per minute, within your organisation’s own limit. Your limit is not a published figure: it is in the X-RateLimit-Limit header of every answer past authentication.

Headers

HeaderWhat it says
X-RateLimit-LimitRequests allowed per minute by the limit that applied to this call: the key's, or the organisation's when that one was reached first.
X-RateLimit-RemainingRequests left in that limit after this one, rounded down.
X-RateLimit-ResetSeconds until the limit is full again. A duration, not a timestamp.

They come with every answer past authentication, including errors, and not with a 401.

When you are told to slow down

Four different budgets answer 429, and they are not retried the same way. RATE_LIMITED sends Retry-After in seconds. QUOTA_EXCEEDED does not clear by waiting: the allowance refills when more is bought.

CodeStatusMeaningRetrying
CONCURRENCY_LIMIT429Too many simultaneous requests.Safe to retry
FLOW_GRANT_EXHAUSTED429That flow grant has no uses left.Do not retry automatically
FLOW_RUN_LIMIT_REACHED429Too many flow runs are active.Safe to retry
OCR_QUEUE_SATURATED429The document queue is full. Retry shortly.Safe to retry
QUOTA_EXCEEDED429The allowance for this window is spent.Do not retry automatically
RATE_LIMITED429Too many requests.Safe to retry

Written from FinSour platform API contract (finsour development at 80e8b83bf), last changed . OpenAPI document at commit 436311c32, API version 2026-11-01. The beta is served from the staging gateway.