How Claude Code usage limits work

If you use Claude Code seriously, you've hit the wall: a task stops mid-flight because you've run out of usage. Here's how Anthropic's limits actually work, and how to see them coming.

Two windows: 5-hour and weekly

Claude Code meters usage against a rolling 5-hour window and a 7-day (weekly) window. The 5-hour window is the one that bites during a long session; the weekly window is the one that decides whether you make it to the end of the week. Each resets on its own clock, so you can be fine on one and nearly out on the other.

The separate Sonnet weekly

On some plans there's an additional weekly limit specific to Sonnet, tracked on top of the all-models weekly. That means a heavy Sonnet week can hit a cap even when your overall weekly still has room, which is why it's worth watching the Sonnet line separately.

How resets work

Limits don't refill all at once at a fixed midnight. The 5-hour window rolls continuously, and the weekly window resets on a 7-day cycle tied to your usage. The practical question is never "what's my limit" but "at my current pace, will I make it to the next reset?"

Team and Enterprise plans

On Team and Enterprise plans the meaningful limit is a shared pool across your organisation. An individual app can only see your own Code allowance, not the pooled total, because Anthropic doesn't expose pooled usage to third-party tools. Treat per-user numbers as your slice, not the whole.

Forecasting it instead of guessing

The useful move is to project your burn rate from your real history and see where you'll land by the reset, before you get there. That's exactly what ReserveGauge does: it reads your live Claude Code (and Codex) limits on your Mac, forecasts your week, and tells you if you're on track or about to run dry, all locally, with no account.