Limits & errors
| Limit | Default | Scope |
|---|---|---|
| Requests | 50,000 / hour (sliding window) | per API key, across all your servers |
| WebSockets | 3 simultaneous connections | per API key |
| Topics per WebSocket | 200 | per connection |
| Message size (client → server) | 4 KB | per message |
What counts as a request
- Every REST call, except
GET /v1/usage. - Opening a WebSocket: 1.
- Each
submessage: 1, regardless of how many topics it carries. - Data pushed to you over WebSocket is free. Streaming is far cheaper than polling: one connection following 100 events costs a handful of requests per hour.
Headers
Every /v1 response includes:
RateLimit-Limit: 50000
RateLimit-Remaining: 49231
RateLimit-Reset: 1436 # seconds until the hourly bucket rolls over
When you run out you get 429 with Retry-After (seconds). On WebSocket a sub over the limit answers {"t":"err","code":"rate_limited","retryAfterSec":…} and the connection stays open.
The 4th simultaneous WebSocket is refused at the handshake with HTTP 429. A connection that stops answering pings is dropped after ~90 s and frees its slot.
Check your budget at any time (free): GET /v1/usage.
Error format
{ "error": { "code": "not_found", "message": "Event not found (or finished more than 48h ago)." },
"meta": { "ts": "…", "requestId": "…" } }
| HTTP | code | Meaning |
|---|---|---|
| 400 | bad_request |
Invalid parameter or body |
| 401 | unauthorized |
Missing or unknown key |
| 403 | forbidden |
Key revoked |
| 404 | not_found |
Unknown route/event, or no data on that channel yet |
| 429 | rate_limited |
Limit reached, see Retry-After |
| 500 | internal |
Our fault. Send us the requestId. |
WebSocket close codes
| Code | Meaning |
|---|---|
| 1001 | Server restarting, reconnect with backoff |
| 4001 | Key revoked |