For AI agents: the documentation index is /docs/llms.txt, and this page is available as Markdown.
Rate limits
Limits are per team, set by your plan. Three things are limited: requests per minute, requests in flight at once, and owner research runs at once.
Limits by plan
| Plan | Requests a minute | Requests at once | Research runs at once |
|---|---|---|---|
| Free | 30 | 2 | 5 |
| Starter | 300 | 10 | 25 |
| Growth | 1,000 | 50 | 50 |
| Pro | 2,000 | 100 | 100 |
| Enterprise | Talk to us |
Every request counts toward the minute, free ones included. A replay of a saved answer for the same Idempotency-Key doesn't count; its RateLimit-Remaining is what the minute has used so far.
Headers
Every response carries:
| Header | Means |
|---|---|
RateLimit-Limit |
Requests allowed in the current window |
RateLimit-Remaining |
Requests left in it |
RateLimit-Reset |
Seconds until the window resets |
Retry-After |
On a 429 only: seconds to wait before trying again |
When you hit a limit
You get 429 with rate_limited. Its reason says which limit: requests (too many this minute) or concurrency (too many at once). The body has retry_after in seconds and an upgrade_url, and the response has a Retry-After header:
{
"error": {
"code": "rate_limited",
"reason": "requests",
"message": "Too many requests. Slow down and retry.",
"retryable": true,
"docs_url": "https://westkite.com/docs/api/errors#rate_limited",
"retry_after": 12,
"upgrade_url": "https://app.westkite.com/developers/billing"
},
"request_id": "req_41c9…"
}
Wait retry_after seconds, then retry. Back off further if it happens again.
Research queues instead of failing
Owner research runs over your plan's number at once don't fail. They're accepted with 202 as queued and start as others finish, on your next poll or within a few minutes. When research is busy across every team, teams share the capacity fairly.