Skip to content

Rate Limits

HeyX applies rate limits per workspace. Public API business requests and MCP business tool calls share these limits, usage accounting, and scheduling. MCP protocol discovery calls such as initialize and tools/list are not counted as business requests.

  • 5 requests per second per workspace.
  • 150 requests per minute per workspace.
  • Requests above these limits are rejected and should be retried later.
  • Honor Retry-After and use backoff. Do not automatically retry a non-idempotent mutation merely because it received a timeout or connection error; processing may have continued.
  • 2 fast requests per second per workspace.
  • 60 fast requests per minute per workspace.
  • 1,800 fast requests per hour per workspace.
  • Requests above these limits are not rejected by the soft limit, but they are handled after fast requests when capacity is constrained.
  • Requests above the soft limits may be delayed. If one cannot start within 30 seconds, HeyX returns a retryable timeout response. In practice, this should be rare.

HeyX advertises the hard rate-limit policy on every public API response. Once a workspace key is authenticated, successful responses and 429 responses carry the current workspace counter state:

  • RateLimit-Limit: 5;w=1, 150;w=60
  • RateLimit-Remaining: <per-second-remaining>, <per-minute-remaining>
  • RateLimit-Reset: <seconds-until-second-window-resets>, <seconds-until-minute-window-resets>

When a request is rejected by a hard limit, HeyX also returns:

  • Retry-After: <seconds>

The RateLimit-* headers describe the hard per-second and per-minute limits only. The soft fast-lane limits are internal scheduling limits and are not exposed as response headers.

Authentication, unknown-path, and unsupported-method errors occur before a workspace request is admitted. Their headers therefore advertise the policy with unconsumed counters rather than exposing another workspace’s state.

  • Every workspace plan includes 100,000 API requests per month.
  • Additional usage is billed at EUR 1.00 per extra 10,000 requests rounded up.