Rate limits

Guardian Logs enforces rate limits at both the per-credential and per-organization level, so one noisy token, key, or tenant cannot starve the platform. Ingest and the read API have separate budgets.

Limits

Ingest is counted in events, not requests — a 500-event batch consumes 500 of the budget, so large batches cannot bypass the limit. The read API is counted in requests (one per call).

All limits are fixed 60-second windows and are configurable per environment:

When you exceed a limit

The request is rejected with 429 and nothing is consumed:

{
  "error": {
    "code": "rate_limited",
    "message": "Rate limit exceeded.",
    "retryAfterSec": 12,
    "scope": "token"
  }
}

• scope is "token" (your credential's budget) or "org" (the whole

organization's budget).

• Wait retryAfterSec seconds, then retry.

Handling 429

• Back off and retry after retryAfterSec (with jitter).
• For ingest, batch events to reduce request overhead — but remember the

budget is per-event, so batching does not raise your event allowance.

• Reuse the same idempotency keys on retry so re-sends are

no-ops.

See also: errors.md · idempotency.md · ingest.md.