A cap on how many requests you can send in a time window, protecting the service (and your balance) from runaway loops. OMA enforces per-key limits; see /docs#rate-limits.