[guides] claude code
Claude Code usage limits: the 5-hour and weekly windows
Claude Code on a Pro or Max plan doesn’t bill per token. It draws from two limits that run on clocks: a 5-hour session window and a weekly one. Here is how they work, what eats into them, and how to see what you have left.
Updated 7 min read
Every paid Claude plan meters usage on two windows at once. The session limit resets every five hours. The weekly limit resets once a week, at a fixed time assigned to your account, and covers every model.
The two stack. A quiet week doesn’t lift the 5-hour cap, and staying under every 5-hour cap doesn’t protect the weekly one. Long days of agent work usually hit the session limit first; steady heavy use across the week hits the weekly one.
Both are shared across Claude Code, claude.ai, the Claude desktop app and Claude in your IDE. An afternoon of long chats on claude.ai leaves less for the terminal.
Anthropic describes the plans relative to each other, not in tokens:
| [plan] | [price] | [usage] |
|---|---|---|
Pro | $20 a month | The baseline session and weekly limits. |
Max 5x | $100 a month | 5× Pro’s usage per session. |
Max 20x | $200 a month | 20× Pro’s usage per session. |
Team and Enterprise | Per seat | A per-seat allowance on a 5-hour and a weekly window, sized by seat tier. |
How far that goes depends on what you ask. A message costs more as the conversation grows, because the whole context goes with it. A bigger model, a larger codebase and attached files all cost more per turn.
- In Claude Code: run
/usage. It shows your plan’s usage bars and a breakdown of the last 24 hours and 7 days. The breakdown counts sessions on this machine only. - On the web: claude.ai › Settings › Usage shows both windows and when each resets.
Both tell you how much you’ve used. Neither tells you whether you’ll make it to the reset, which takes comparing usage with how much of the window has passed. That’s the pace Burnstate keeps in your menu bar. To work it out by hand, use the pace calculator, and the reset calculator for the exact reset time.
Claude Code stops with a message naming the limit and when it resets, like 5-hour limit reached - resets 3pm. From there you have three options:
- Wait for the reset. Recent versions of Claude Code can wait and pick the task up again once the window resets.
- Turn on usage credits. On Pro and Max, usage credits (once called extra usage) keep you going past a limit at standard API rates. Turn them on in Settings › Usage, with a monthly spend cap. On Team and Enterprise, an admin sets them up.
- Move up a plan. If you hit the session limit most days, Max 5x or 20x costs less than steady usage credits. The cost calculator shows what your own logs would cost at API prices.
- Clear between tasks.
/clearstarts fresh, so the next task doesn’t carry the last one’s context./compactshrinks a long session you want to keep. - Match the model to the job. Use the biggest model for planning and hard bugs, and a smaller one for routine edits.
- Watch parallel agents. Subagents and several sessions at once multiply the burn rate. They finish sooner, and they empty the window sooner too.
- Check the pace, not the total. Half used with an hour left is fine. Half used ten minutes in is not.
Burnstate reads your session, weekly and per-model limits with the Claude Code login already on your Mac, or from claude.ai when Claude Code isn’t signed in. Each limit gets a bar with a white tick for the clock, a status, and a forecast of when you’d run out at your current pace. See Connect providers and Limits and pace.
Plans and limits change often. These are the pages this guide was checked against.
[burnstate]
See every limit against the clock, in your menu bar.
Burnstate reads your Claude Code, Codex, Cursor and Vercel limits, marks how much of each window has passed, and tells you when you’d run out at your current pace.