Rate limits

Each team has a request allowance per minute that follows what the team pays for:

Bursts are fine. You can spend a whole minute's allowance at once; what counts is the average over a short window. Requests we refuse do not count against you.

Headers#

Every answer includes:

• x-ratelimit-limit: the requests a minute your team's tier allows.
• x-ratelimit-remaining: what is left right now.

When you hit the limit#

The answer is 429 rate_limited, which is retryable and free, with a retry-after header holding the exact number of seconds to wait:

{
  "ok": false,
  "error": { "code": "rate_limited", "message": "Too many requests for this team's plan; wait for Retry-After.", "retryable": true },
  "credits_charged": 0,
  "request_id": "req_0123456789abcdef01234567"
}

Wait for retry-after, then continue. The retry helper already does this. Do not spread a burst over many keys: the limit is per team, not per key.

Batching#

A batch of up to 100 requests runs concurrently on our side and saves round trips, but each item still counts toward the limit.

More headroom#

Plans raise the limit; see Credits and pricing. For sustained volume above the top tier, write to [email protected].