PromptFloe Developer Docs
Reference

Rate limits

PromptFloe enforces two layers of limits: per-second API rate limits (to protect the service) and per-day skill quotas (tied to your plan). Both surface in response headers.

#API rate limits (per second)

FieldTypeDescription
Free5 req/sPer workspace, across all endpoints.
Basic20 req/s
Pro60 req/s
Max200 req/s
EnterprisecustomNegotiated per contract.

#Daily skill quotas

Critique and augmenter runs each have separate counters. Reset at 00:00 UTC. See Billing for the per-tier table.

#Headers

Every response includes:

FieldTypeDescription
X-RateLimit-Limit-RequestsnumberPer-second API request budget.
X-RateLimit-Remaining-RequestsnumberRequests left in the current 1s window.
X-RateLimit-Reset-RequestsmsMilliseconds until the window resets.
X-RateLimit-Limit-CritiquenumberDaily critique-skill cap.
X-RateLimit-Remaining-CritiquenumberCritique runs left today.
X-RateLimit-Limit-AugmenternumberDaily augmenter cap.
X-RateLimit-Remaining-AugmenternumberAugmenter runs left today.
X-RateLimit-Reset-DailyISO 8601When daily counters reset.
Retry-AftersecondsOn 429, suggested seconds to wait before retrying.

#Example 429 response

http
HTTP/1.1 429 Too Many Requests
Retry-After: 2
X-RateLimit-Limit-Requests: 20
X-RateLimit-Remaining-Requests: 0
X-RateLimit-Reset-Requests: 1000

{
  "error": {
    "code": "rate_limited",
    "message": "Per-second rate limit exceeded.",
    "retryAfterMs": 2000
  }
}

#Best practices

  • Read the headers — log X-RateLimit-Remaining-* in your APM. You'll see throttling coming before it hurts.
  • Honor Retry-After on 429s. The SDK does this automatically.
  • Batch where possible — many small calls eat the per-second budget faster than fewer larger ones. Use list endpoints with appropriate limit instead of polling individual ids.
  • Use webhooks for long-running work instead of polling.

#Where to go next

PromptFloe developer docs