[docs]
Limits and pace
Every limit runs on a window, like five hours or a week. Burnstate shows how much of the limit you’ve used and how much of its window has passed, so you can tell whether you’ll make it to the reset. To do the math by hand, see how to pace AI coding limits.
Each provider row on the home screen has a bar of small cells. The filled cells are what you’ve used. The white tick is the clock: how far through the window you are.
- Filled cells behind the tick: you’re using less than the clock allows. The row says how far behind, like 24% behind.
- Filled cells past the tick: you’re burning faster than the window allows, like 11% ahead. Keep that pace and you run out before the reset.
Under the bar, the row names the limit it shows (with +2 if the provider has two more) and when it resets.
Each limit gets a status from what’s left and how fast it’s going. The home screen sorts providers by it, so the one that needs you is on top.
| [status] | [when] |
|---|---|
on track | At least 30% left, and no more than 5 points ahead of the clock. |
burning fast | Less than 30% left, or more than 5 points ahead of the clock. |
running out | Less than 15% left, or more than 20 points ahead of the clock. |
Home groups them as [needs attention], [on track] and [paused or stale]. A balance with no reset, such as credits, says running low where a limit would say burning fast.
Select a row to open the provider. Its first section is the limit that matters most right now:
- A big number with what’s left or used, the limit itself, and the status.
- The bar with its tick, the clock in percent, and how far ahead or behind it you are.
- When you’re burning fast or running out, a forecast like runs out 15:04, 38m before reset.
Under it, as far as the provider reports them:
| [section] | [shows] |
|---|---|
[limits] | Every other window, each with its own bar, status and reset. |
[balances] | Credits that add up to the main limit. |
[today] | Tokens, requests, cost at API prices, and the model you used most. |
[models] | How today’s tokens split between models. |
[7 days] | Tokens per day for the last week. |
[cli] | The installed and latest version of the tool’s CLI, and how to update it. |
Select ‹ providers or press esc to go back.
Tokens and cost come from the logs your tools keep on your Mac. Cost uses published per-token API prices, including cache reads and writes. A subscription isn’t billed per token, so read it as what that work would cost at API rates.
Bars and percents show what’s left by default. Switch them to what you’ve used, or pin which of a provider’s limits the home row and the menu bar follow, in settings.
With many providers connected, press ⌘F on the home screen and type to filter them. Return opens the first match, and esc clears the search.