Platform

Rate limits

Limits protect shared infrastructure while giving authenticated integrations predictable throughput.

Anonymous users

  • Create: 20 requests / hour / IP
  • Status & download: 120 requests / hour / IP
  • In production, a resolvable client IP is required for anonymous access

Authenticated API keys

  • Default create window: 60 / minute (configurable per key)
  • Daily create quota: 2000 / UTC day
  • Max concurrent PENDING/PROCESSING tasks: 5
  • Status/download limit is roughly 6× the create window limit (minimum 120)

Burst and daily limits

Window limits smooth short bursts. Daily quotas cap sustained create volume. Hitting either returns 429 with code: rate_limit_exceeded.

Response headers

  • X-RateLimit-Limit
  • X-RateLimit-Remaining
  • X-RateLimit-Daily-Remaining — API key creates only
  • Retry-After — seconds to wait on 429

429 handling and backoff

Recommended strategy

Honor Retry-After. Use exponential backoff with jitter for create retries, and poll status every 1–2s (increasing to ~8–10s). Always send Idempotency-Key on create retries.

Best practices

  • Separate create QPS metrics from poll QPS
  • Cap client concurrency to stay under the concurrent task limit
  • Cache COMPLETED status locally — do not re-poll finished tasks
  • Upgrade to a higher plan before saturating daily quotas in production

Max upload size is 20 MB for all tiers unless Enterprise overrides apply.