# Rate limits

> Limits are per API key and per endpoint. Over a limit ICON answers 429 with Retry-After.

## Defaults

| Endpoint | Per minute | Burst |
| --- | --- | --- |
| Search: `POST /api/v1/engine/search` | 60 | 10 |
| Results polling: `GET /api/v1/engine/results/{icon_lead_id}` | 600 | 600 |
| Submit: `POST /api/v1/engine/leads` | 120 | 120 |
| Duplicate pre-check: `POST /api/v1/engine/dupe-check` | 30 | 10 |

Each key and endpoint has its own budget. It refills continuously (per minute ÷ 60 each second) and holds at most the burst, so a key can send a burst at once and then its steady rate. Treat the numbers as your ceiling, not a guarantee of capacity. ICON can set other limits on a key; ask your ICON contact.

## Over a limit

A call over a limit may be refused:

```http
HTTP/1.1 429 Too Many Requests
Retry-After: 2
Content-Type: application/json

{
  "code": "RATE_LIMITED",
  "status": "failed",
  "reason": "Too many requests",
  "retryable": true,
  "retry_after_seconds": 2
}
```

Wait the `Retry-After` seconds, then retry. See [Errors and retries](https://developers.iconroute.io/errors/).

## Staying under

- Poll results no faster than `poll_after_ms` and stop at `processing_done`.
- Search once per lead. Submit only results the consumer chose.
- Use one key per integration so one busy integration cannot slow another.