Core concepts
Read your remaining quota from response headers, and back off before you hit 429.
How limits work
Rate limits apply per application, per API. They are enforced with a token bucket: you have a steady allowance per minute, plus a small burst buffer for short spikes.
If you exceed the limit, the API returns 429 Too Many Requests and the request is not processed.
Read the headers
Every response includes your current usage:
| Header | What it tells you |
|---|---|
X-RateLimit-Limit | Your allowance for the current window |
X-RateLimit-Remaining | How many requests you have left |
X-RateLimit-Reset | When the window resets, as a Unix timestamp |
Retry-After | On 429 responses, how long to wait in seconds |
curl -i "https://api.example.com/v1/orders" \
-H "Authorization: Bearer $API_KEY"
X-RateLimit-Limit: 600
X-RateLimit-Remaining: 597
X-RateLimit-Reset: 1727712000
Log these headers in production. A shrinking X-RateLimit-Remaining is the earliest warning that your traffic pattern needs work.
Handle 429s correctly
When you get a 429:
- Read the
Retry-Afterheader and wait that long. - Retry the request with exponential backoff as a fallback.
- Never retry in a tight loop. It burns quota and delays your own recovery.
Idempotency and retries
If a retried request creates something, send an idempotency key so a timeout can't double the write. See Idempotent requests.
Design around the limit
- Batch reads where the API allows it, instead of polling one resource at a time.
- Cache responses that rarely change.
- Spread scheduled jobs across the window instead of firing them all at midnight.
Need a higher limit? Search the guides, or contact the team that runs these APIs and include your application name and typical request volume.