Rate Limits
Rate limits protect platform stability. When you exceed limits you’ll receive 429 responses. Implement backoff and concurrency control to stay within your quota.
What to do on 429
Back off and retry
Use exponential backoff with jitter to avoid synchronized retries.
Reduce concurrency
Cap parallel requests per tenant/project and smooth traffic spikes.
Limits by plan
| Plan | Req/sec | Req/min | Req/hour | Burst |
|---|---|---|---|---|
| Starter | 10 | 500 | 10,000 | 20 |
| Business | 100 | 5,000 | 100,000 | 200 |
| Enterprise | 1,000 | 50,000 | — | 2,000 |
Limits vary by tenant, environment, endpoint, and risk posture. Your responses include rate limit headers (see below).
A safe default algorithm
- Retry only retryable requests (e.g., GETs, or POSTs with idempotency keys).
- Use exponential backoff: 250ms → 500ms → 1s → 2s (cap at ~10s).
- Add jitter (randomize delay) to avoid thundering herds.
- Stop after a small number of attempts and surface the failure.