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

# Rate Limits — provider limit visibility

> /ratelimit shows the provider's rate limits, parsed passively from the x-ratelimit-* response headers — remaining requests and tokens, % used, and time until reset.

The **`/ratelimit`** command (alias **`/limits`**) shows the current state of your LLM provider's rate limits, **without making any extra call**: ChatCLI reads the `x-ratelimit-*` headers the provider already returns on every response and keeps a per-provider snapshot.

<Info>
  Capture is **passive and central**: an observer on the auth path intercepts the HTTP responses of **all providers** that send the `x-ratelimit-*` header family (OpenAI, OpenRouter, Anthropic-compatible endpoints, etc.). No environment variable is required.
</Info>

***

## Usage

```text theme={"system"}
> /ratelimit
  Rate Limits (from provider headers)
  ─────────────────────────────────────────
  OPENAI
    requests  4987 / 5000   (0% used, resets in 12s)
    tokens    789012 / 800000   (1% used, resets in 48s)
```

Before the first request there is no data:

```text theme={"system"}
> /ratelimit
  Rate Limits (from provider headers)
  ─────────────────────────────────────────
  No rate-limit data captured yet. Make a request first.
```

***

## What is shown

For each provider that reported limits, ChatCLI shows up to two buckets:

| Bucket     | Fields                                           |
| ---------- | ------------------------------------------------ |
| `requests` | remaining / limit, **% used**, **resets in N s** |
| `tokens`   | remaining / limit, **% used**, **resets in N s** |

The "resets in" value is adjusted for the time elapsed since capture, so it reflects the real remaining time at the moment you run the command.

<Tip>
  Useful in pipelines and long sessions to anticipate throttling: if `requests` or `tokens` is near 0% remaining, wait for the reset (or switch provider with `/switch`) before firing a heavy `/agent`.
</Tip>

***

## See also

* [Provider Fallback](/features/provider-fallback) — automatic switch when a provider fails/saturates
* [Session Cost](/features/cost-tracking) — `/cost`
* [Command Reference](/reference/command-reference)