Where it is raised
- Checks: 8 routes
- Metrics: 1 route
- Status Pages: 5 routes
- Notification Channels: 2 routes
- Incidents: 2 routes
- Maintenance: 1 route
- Errors: 5 routes
- Agents: 3 routes
- Keyed MCP endpoint, read tools
- Check and result exports, which answer Export rate limit exceeded from their own per-account budget
What triggers it
Two budgets answer with this one code. Every key has a read bucket of 120 requests a minute, and a Free account also shares 60 requests a minute across all its keys. The Free account budget uses the shared error response, so that reply carries a docs link. A read request is a GET or HEAD.
Example response
GET /api/v1/checks
HTTP 429
Retry-After: <seconds>
{
"error": "Rate limit exceeded. Try again shortly.",
"code": "RU-2002"
}Common causes
- Polling in a tight loop instead of respecting Retry-After.
- Many workers sharing one API key.
- Three integrations on a Free account, each with its own key, all reading from the one 60-a-minute account budget.
How to fix it
- Honor the Retry-After header before retrying.
- Cache reads that don't need to be fresh every second.
- Give independent systems their own API keys so one noisy consumer doesn't starve the rest.