RATE LIMITS

Rate limits and quotas

Two independent limits protect the API: a per-second rate limit and a daily or monthly quota. There is never any overage billing; requests simply stop until the limit resets.

How limits work

Rate limit

A sustained number of requests per second per key, with a short burst allowance on top (a token bucket). Exceeding it returns 429 rateLimited with Retry-After.

Quota

The total requests your plan includes. Every plan, Free included, resets at the start of each UTC calendar month. Exceeding it returns 429 quotaExceeded.

Plan limits

Each plan’s credits and entitlements are listed on Plans, and your current limits are shown in the console.

Need more? Compare plans or upgrade from the console; the new limits apply within 60 seconds.

Response headers

Authenticated responses tell you where you stand, so you can slow down before you hit a limit:

HTTP/1.1 200 OK
X-RateLimit-Limit: 5
X-RateLimit-Remaining: 9
X-Quota-Limit: 20000
X-Quota-Remaining: 18542

Rate-limit headers are sent once the key is authenticated; quota headers are sent once the request passes the rate limit. All of them are readable from browser JavaScript (CORS-exposed).

When you hit a limit

• rateLimited: wait the number of seconds in Retry-After, then retry.
• quotaExceeded: don't retry; it won't succeed until X-Quota-Reset. Stop and alert, or upgrade.

A copy-paste retry helper is in Retrying safely, and the SDKs handle it automatically.

Staying under the limit

• Request bigger pages. One call with page_size=100 costs the same as one with 10.
• Poll for new articles only. Use indexed_from with sort=indexed_desc instead of re-running the same search. See Keep a RAG index fresh.
• Cache results for repeat queries in your own app.
• Watch X-Quota-Remaining and alert before it reaches zero. Usage charts in the console show the trend.