Skip to main contentRate limits
Rate limits can vary by product and deployment. Build clients that are resilient even when limits are not explicitly
documented.
Best practices
- Prefer exponential backoff with jitter for retries.
- Retry only when safe:
- ✅
GET requests are usually safe
- ⚠️
POST may not be safe unless the endpoint is idempotent (or you use a request ID)
- Treat
429 and 503 as retryable by default, using backoff.
- Put a cap on retries and fail fast for clearly invalid requests (
400, 401).
Example backoff schedule
Try 1s → 2s → 4s → 8s (plus small random jitter), then stop and surface an error. 