Rate Limits
Plan your request throughput and handle rate limiting.
Rate limits are enforced per endpoint. See each endpoint's reference page for its sustained rate and burst.
#How limits are enforced
Each endpoint has two caps:
- Sustained rate — requests per minute, smoothed over time.
- Burst — short-window allowance over the sustained rate.
Exceeding either returns 429.
Agentic Search also limits pending/running jobs to 2 per API key and 3 per account. Its concurrency rejection uses the API error envelope with error.type: "rate_limited"; the gateway rate-limit response below uses a different format.
#429 response shape
HTTP/1.1 429 Too Many Requests
{"message":"Too Many Requests"}We do not return a Retry-After header and recommend jittered exponential backoff.
#Catch-all fallback
Requests to paths not in the explicit endpoint list (typos, missing trailing slash, deprecated paths) fall through to a catch-all with a much tighter limit:
| Limit | Value |
|---|---|
| Sustained rate | 1 request / minute |
| Burst | 2 requests |
/enrich/email (no slash) hits the catch-all; /enrich/email/ hits the explicit endpoint.