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

# Rate limits

> How the gateway paces requests, and how to handle 429s.

The gateway rate-limits per API key. Limits are split into two classes so that
reads never starve your ability to place and cancel orders:

| Class     | Endpoints                                              | Notes                            |
| --------- | ------------------------------------------------------ | -------------------------------- |
| **Read**  | `GET` market data, positions, fills, risk stats        | Higher ceiling                   |
| **Write** | `POST`/`PUT`/`DELETE` — orders, cancels, limit changes | Lower ceiling; the critical path |

Mithril also absorbs each **venue's** own limits server-side (Polymarket's
opaque Cloudflare pacing, Kalshi's tiered token buckets), so you code against
one predictable ceiling instead of two moving ones.

## When you're limited

A throttled request returns `429` with the standard [error envelope](/concepts/errors):

```json theme={null}
{ "error": { "code": "rate_limited", "message": "rate limit exceeded", "request_id": "req_..." } }
```

Back off and retry. A simple exponential backoff with jitter is enough:

```python theme={null}
import time, random

def with_backoff(fn, tries=5):
    for i in range(tries):
        resp = fn()
        if resp.status_code != 429:
            return resp
        time.sleep(min(2 ** i, 8) + random.random())
    return resp
```

<Tip>
  Prioritize cancels over new orders in your own client, and prefer
  [worked orders](/guides/smart-execution) over hand-rolled cancel/replace
  loops — Mithril's execution engine paces child orders within venue limits for
  you.
</Tip>

<Note>
  Need higher throughput for a market-making strategy?
  [Talk to us](https://calendly.com/thekarlg) — limits are lifted per desk.
</Note>