Skip to main content
Errors are returned as JSON with the standard RFC 7807 problem-details shape:

HTTP status codes

Rate limits

The API limits each API key to 100 requests per 60-second window (with a small queue of 10 requests when the limit is reached). When you exceed the limit, you’ll get 429 Too Many Requests. Three response headers communicate the limit on every request: If you’re hitting the limit consistently, contact support — limits are config-driven per deployment and can be raised. Limits apply per API key, so distinct keys (e.g., per-environment) get independent budgets.

Retries

Retry 5xx and 429 responses with exponential backoff, respecting Retry-After when present. Do not retry 4xx responses other than 429. POST endpoints are not currently idempotent. A retried POST /agent/{id}/start will queue a second Run — design your retry logic to check before retrying (e.g., poll GET /agent/{id}/runs for a recent in-flight Run before issuing another start).