Rate limits

The platform enforces two rate limits in parallel:

• Per-tenant — every company has a request budget shared by all of its keys and OAuth grants.
• Per-key — each key has its own budget that cannot exceed the tenant budget.

Both budgets refill on a sliding window.

Headers

Every response includes:

• X-RateLimit-Limit — the smaller of (tenant, key) limit.
• X-RateLimit-Remaining — requests left in the current window.
• X-RateLimit-Reset — unix epoch when the window rolls over.
• Retry-After — present on 429s; seconds to wait.

Backoff strategy

The official SDKs implement exponential backoff with jitter automatically. For custom clients:

1. On 429, sleep Retry-After seconds (with at least 100ms of jitter).
2. On 5xx, retry idempotent requests with backoff (max 3 attempts).
3. Never retry non-idempotent writes without an Idempotency-Key.

Bulk endpoints

Use the bulk variant whenever you need to write more than ~10 items. Bulk endpoints count as one request against your budget regardless of batch size.