# Usage & Limits

import { Callout } from "nextra/components";

Pane tracks how much each agent costs and where you stand against provider rate limits.

## The Usage & Limits screen

Open **Settings → Usage & Limits** to see:

- **Tokens**: input and output tokens per agent CLI (Claude Code, Codex, Cursor Agent).
- **Cost**: estimated spend based on model prices fetched from OpenRouter.
- **Rate limits**: current usage against your provider's rate-limit tiers, per agent.

The screen refreshes on a four-hour scan cycle. Pull-to-refresh manually if you need the latest numbers immediately. You can also share or download the usage summary from the screen.

## Per-pane cost attribution

Every pane tracks the tokens and cost of the agent work that happened inside it. This lets you see which features or tasks are expensive.

### In the UI

Per-pane cost shows alongside pane metadata in the sidebar and pane details.

### From the CLI

```bash
runpane panes cost --json
```

Returns cost attribution for all panes, or filter to one:

```bash
runpane panes cost --pane <pane-id> --json
```

The report covers the last 30 days. Cost is attributed from session-level token counts. Archived panes retain their cost data within that window.

<Callout type="info">
  Cost tracking depends on model pricing data from OpenRouter. If a model's pricing isn't available, the token count is still recorded but the dollar estimate may be missing.
</Callout>

## Codex usage

If you're signed in to Codex, a **Codex Usage** tab appears in Settings showing your Codex-specific usage and limits. This is separate from the general Usage & Limits screen and pulls data from your Codex account.

## Related

- [AI Providers](/docs/providers): configure which models and providers your agents use
- [runpane CLI](/docs/runpane-cli): full command reference including `panes cost`
- [Settings](/docs/settings): other app-level settings