> ## Documentation Index
> Fetch the complete documentation index at: https://docs.augment.market/llms.txt
> Use this file to discover all available pages before exploring further.

# Rate limits

> Usage tiers, quotas, and throttling behavior

Every API key is assigned to a usage tier at the time it is minted. The tier
is not something you can change yourself — contact Augment to change it.

## Tiers

| Tier        | Daily quota  | Sustained rate | Burst |
| ----------- | ------------ | -------------- | ----- |
| `standard`  | 10,000 / day | 10 requests/s  | 20    |
| `throttled` | 1,000 / day  | 2 requests/s   | 5     |

All three limits are enforced independently. The sustained rate is the
average you can hold indefinitely; burst is the depth of the token bucket
absorbing short spikes above it. Exceeding any of the three returns the same
error:

```json theme={null}
{
  "error": "Too many requests, please try again later",
  "code": "rate_limited"
}
```

with HTTP status `429` — whether you tripped the sustained rate, the burst,
or the daily quota.

There is no `Retry-After` header on this response today. Back off and retry
using your own client-side backoff strategy; do not rely on a header telling
you when to retry.

## Service unavailable

If the entire External API surface is switched off (a kill switch, not a
per-client rejection), every request — including ones from valid, in-quota
keys — returns:

```json theme={null}
{
  "error": "Service temporarily unavailable, please try again later",
  "code": "service_unavailable"
}
```

with HTTP status `503`. This is distinct from `rate_limited`: it means the
whole API is down, not that your key hit a limit.