Claude Code
Claude Code usage, without /usage.
Claude Code caps you twice: a rolling 5 hour session and a weekly cap, with a separate weekly cap for some models. Hit either and Claude stops mid task until the window resets. The CLI knows the numbers and keeps them behind a slash command.
The built-in way. Ask the CLI.
Run /usage inside a Claude Code session.
It shows the session bar, the weekly bar and their reset times. That works, once, and it costs a terminal switch every time you want to look.
What Notchlet shows. One ring per window.
- 5h session
- A rolling window that starts with your first message and resets five hours later.
- Weekly
- Every model together, reset once a week.
- Per model
- A second weekly cap for some models, shown as its own ring.
Hover the notch and each window is a ring with the percentage left, the reset time and a verdict on your burn rate. Click a ring for history: cost and tokens per day, and every model you used.
How it signs in. With the login you already have.
Notchlet reads the token Claude Code keeps in your keychain, with its credentials file as fallback, and calls the endpoint behind /usage. It reads the item the way Claude Code writes it, so there is no password prompt. When the token expires while Claude Code is idle, Notchlet refreshes it the way a second Claude Code process would, with the same locks and the same write-back, so Claude Code stays signed in.
Questions. Short answers.
- How do I check my Claude Code usage?
- Run /usage in a Claude Code session, or install Notchlet and hover the notch. Both read the same numbers.
- Does Notchlet spend Claude Code usage to check it?
- No. The usage endpoint reports the windows. It does not start a conversation.
- Do I need to paste an API key?
- No. Notchlet uses the login Claude Code already has. A Console API key would not show plan usage anyway.