Usage limits
Claude Code usage limits: 5-hour window, weekly cap, and resets
Last reviewed August 2026.
Short answer: Claude Code usage limits are two clocks at once — a rolling 5-hour window and a weekly cap — both set by your Claude plan. The 5-hour window refills continuously; the weekly cap resets on a fixed 7-day schedule tied to your account. What drains them is tokens: every message, response, file read, and log line. You hit the wall fastest on token-heavy agent sessions. Exact allowances vary by plan and change over time — verify against Anthropic’s current docs — but the shape below is stable as of July 2026.
The two clocks that govern Claude Code
Claude Code on a paid plan does not give you unlimited agent time. Anthropic meters usage with two overlapping ceilings:
- A short rolling 5-hour window that governs burst intensity.
- A longer weekly ceiling that governs how much you can spend across seven days.
Both apply at once. Clearing one does not clear the other. That interaction is what makes “I was fine all morning and suddenly stopped” feel mysterious until you know which clock fired.
This page is the hub for how those limits work. For the window mechanics in depth, see the Claude Code 5-hour limit guide. For Max-tier allowances specifically, see Claude Max limits. If you are on Windows or WSL and want the setup side, start from the Claude Code on Windows hub.
The 5-hour rolling window
The first clock is a rolling five-hour window. Your first message starts a session, and everything you do for the next five hours draws from the same allowance. Heavy back-to-back work — long agent runs, big file edits, repeated test loops — drains that window fast. When it empties, you wait for it to roll over before you can continue at full pace.
Because it is rolling rather than a fixed clock face, it refills continuously through the day rather than snapping back at a single wall-clock time. The dedicated 5-hour limit deep-dive covers session starts, what “rolling” means in practice, and how it interacts with the weekly cap.
The weekly cap
The second clock is the one that catches people mid-week. On top of the rolling window, paid plans carry a weekly ceiling that resets on a fixed schedule. You can stay under the five-hour limit every single day and still hit the weekly cap by Wednesday if your sessions are token-heavy. Once the weekly ceiling is reached, waiting out the 5-hour window will not help — that is a separate clock.
The two limits interact in a way that is easy to miss: a single noisy afternoon — a build that dumps thousands of lines of logs into the context, a directory read that pulls in files you never needed — can quietly eat a meaningful slice of your weekly budget, not just today’s window.
Limits by Claude plan
Allowances scale with your Claude subscription. Anthropic adjusts the exact numbers over time, so treat the table as the relative shape as of July 2026, not a promise of fixed token counts. Confirm against your plan’s current details in Anthropic’s documentation or account UI.
| Claude plan | 5-hour window | Weekly cap |
|---|---|---|
| Pro | Standard paid allowance | Standard weekly ceiling |
| Max (lower multiplier) | Substantially higher than Pro | Substantially higher than Pro |
| Max (higher multiplier) | Highest consumer-tier burst room | Highest consumer-tier weekly ceiling |
| Team / Enterprise | Seat-based / negotiated | Seat-based / negotiated |
Relative scaling as of July 2026; Anthropic may change absolute figures. Max-specific detail is in Claude Max limits.
API keys and the Claude API sit on a different system entirely — request-rate and spend tiers, not the subscription 5-hour/weekly windows. If you are debugging HTTP 429s from the API, see Anthropic rate limits; that is a distinct intent from the subscription caps on this page.
When each limit resets
| Clock | How it resets | What waiting does |
|---|---|---|
| 5-hour window | Rolling — roughly five hours after the session that started it | Clears continuously; often free again the same day |
| Weekly cap | Fixed 7-day cycle tied to your account | Waiting out the 5-hour window does not refill it |
If you are capped, the CLI or product UI usually indicates which clock is blocking you and roughly when it frees up. Knowing which limit fired is the first step, because the fix differs: a spent 5-hour window clears on its own within hours; a spent weekly cap means wait for the weekly reset, upgrade the plan, or spend fewer tokens per session on what remains.
What drains the limit fastest
Usage is token-metered. A short back-and-forth costs little; a long agent run that reads dozens of files, replays a failing test suite, and carries a growing conversation costs a lot. Context is re-processed on each turn, so noisy content you pulled in earlier keeps costing you until the session ends. That is why two developers on the same plan can have wildly different mileage.
The biggest drains tend to be:
- Verbose tool and build output stuffed into context
- The same files re-read across many turns
- Extra sub-agent or exploration context you did not need
Why Windows and WSL developers feel this harder
The tools that help you see this burn and cut it were, as of July 2026, largely built for macOS. If you code on Windows or inside WSL2 — a large share of the Claude Code community — you have historically had almost nothing: no native tray meter, no forecast, no before-and-after view of what fills your context. You find out you are capped when the terminal tells you.
That is the gap Hypermile is built for: a cross-platform usage copilot for Windows (including WSL2), macOS, and Linux — not a Mac-only afterthought. Setup and path details live in the Claude Code on Windows and Claude Code WSL guides.
How to actually see your burn
You do not have to guess. Claude Code writes a transcript of every session to your local machine, and those transcripts contain everything you need to break down where your tokens went — tool output versus file reads versus conversation versus sub-agents.
The free hypermile-audit CLI reads those local transcripts and prints a plain-English report: how much context you processed, which categories dominated, and where the obvious waste is. Nothing is uploaded; it runs entirely on your machine and it is open source.
Once you can see the breakdown, the fixes get concrete. Maybe a large share of your week is compressible log output. Maybe one module got re-read a dozen times. You cannot trim what you cannot measure.
What to do next
Run the audit below to get your own numbers, then download Hypermile that forecasts your cap before you hit it. Dig into the 5-hour window, weekly limit, usage limit reached, Claude Max limits, Claude Pro usage limits, how to check your usage, what counts toward token usage, or Anthropic API rate limits if that is the message you are seeing. Weighing tools? Our honest Headroom alternative write-up covers how the options differ for Windows and WSL users.
Frequently asked questions
What are Claude Code usage limits?
Claude Code on a paid Claude plan is gated by two limits at once: a rolling 5-hour usage window and a weekly ceiling. Both are set by your plan tier. Hitting either one pauses new work until that clock resets or you move to a higher tier.
When does the Claude Code usage limit reset?
The 5-hour window refills on a rolling schedule — roughly five hours after the session that started it — so it clears continuously through the day. The weekly cap resets on a fixed 7-day cycle tied to your account, not on Monday of the calendar week.
Is the Claude Code 5-hour limit the same as the weekly limit?
No. They are separate clocks. You can stay under the 5-hour window every day and still hit the weekly ceiling mid-week if your sessions are token-heavy. Waiting out the 5-hour window will not clear a spent weekly cap.
Do Claude Max plans have higher usage limits?
Yes. As of July 2026, Claude Max tiers carry substantially larger 5-hour and weekly allowances than Pro. Exact figures change as Anthropic tunes them, so confirm against your current plan details. Our Claude Max limits guide covers the Max-specific shape.
Why did I hit my Claude Code usage limit so fast?
Usage is token-metered, not time-metered. Long agent runs, verbose tool output, and files re-read across turns drain both clocks quickly. A single noisy afternoon can consume a disproportionate slice of the weekly budget even when the 5-hour window still looks fine.