Usage and limits
Claudette’s main job beyond the tabs is to show how much of your Claude plan’s limits you’ve used, how fast, and when you’ll run out, without you running /usage. This chapter covers the meters in the header and the charts behind them, what each tab has used, the warnings, what happens when a limit stops a task, and the history Claudette keeps so it can draw all of it.
Usage here means your Claude plan’s limits. If a tab runs under an API key instead, it has no plan limits, and the header shows that key’s estimated spend: see Accounts.
In this chapter
- The usage headerThe strip across the top of the window shows how much of your plan's limits you've used, how fast you're using them, and when you'll hit the limit at that rate.
- The detailed headerThe header can be drawn taller, with charts, so you can keep an eye on usage without opening the Usage window.
- The Usage windowThe Usage window is the header's detail: a larger chart of the session, which tab and which project used the most, the week, and how past sessions and weeks went.
- A tab's tokensEach tab counts the tokens it has used.
- Context and compactingSeparate from your plan's limits, each tab has a context window: how much of the conversation Claude can hold at once.
- The prompt cacheClaude keeps a conversation in a cache for a while after each request that reads or writes it, for five minutes or an hour.
- AlertsClaudette tells you when your session is getting full, so you don't have to watch the header.
- When a limit stops a taskWhen a plan limit stops a task, the tab waits for the reset and carries on by itself.
- Tabs that don't use usage creditsOnce a limit is used up, a tab can stop and wait for the reset rather than spend your usage credits.
- Usage historyClaudette keeps your usage on this machine, in a small database, so the trendline, the charts and the per-tab numbers survive a restart.
- Sharing usage across machinesYour plan's limits belong to your account, so every machine signed in to it sees the same usage.