Claude Code limits, explained
If you use Claude Code on a Pro or Max subscription, your usage is governed by rolling limits rather than a per-token bill. The mechanics are simple once laid out, but they're spread across several docs. Here's the short version, plus how to track where you stand. (Details as of July 2026 — Anthropic adjusts limits over time, so treat /usage as the source of truth.)
The 5-hour session window
Usage is metered in rolling 5-hour sessions: your first message starts a window, and everything you send in the next five hours counts against it. Hit the session cap and you wait for the window to reset. Both the Claude app and Claude Code draw from the same pool — a heavy afternoon of chat also eats your coding budget.
Weekly caps
On top of the session window, subscriptions have weekly caps that reset every seven days: one covering all models, and on Max plans a separate one for Opus. These mostly matter to heavy users — if you regularly hit session limits, the weekly cap is the next ceiling you'll meet.
What the limits actually count
Anthropic doesn't publish exact token quotas, and effective capacity varies with model choice and context size. Practically: Opus consumes your allowance several times faster than Sonnet, and long contexts (big files, long sessions) consume it faster than short ones.
How to see where you stand
/usagein Claude Code — live session and weekly percentages. The authoritative number.- A custom status line can show rate-limit state persistently while you work.
- Claude Code Usage Monitor predicts when you'll hit the limit at your current burn rate.
npx ccusage blocksgroups your local usage into the same 5-hour windows the limit uses, so you can see your historical pattern.
Stretching a limited budget
- Match the model to the task. Mechanical edits and simple questions don't need Opus. Model choice is the single biggest lever.
- Keep contexts small. Start new sessions for new tasks and use
/compact— resending a huge conversation with every turn is what drains windows fastest. - Batch related questions instead of many small turns; each turn re-sends context.
- API key as overflow. If you hit a wall mid-task, switching to pay-per-token API billing for the remainder is often cheaper than a plan upgrade you rarely need.
Tracking usage over time
Limits are about the next five hours; habits show up over weeks. /stats gives you a personal dashboard, ccusage gives you local reports, and if you're curious how your usage compares with friends, ccclub (our project) puts a group on one leaderboard from the same local logs — see all the ways to check usage.
FAQ
How does the Claude Code 5-hour limit work?
Your first message starts a rolling 5-hour session window, and usage within that window counts against a session cap. When you hit it, you wait for the window to reset. Claude app usage and Claude Code usage share the same pool.
How do I check how close I am to my Claude Code limit?
Run /usage inside Claude Code — it shows live session and weekly usage. For continuous visibility, use a custom status line or a real-time monitor like Claude Code Usage Monitor.
Why am I hitting Claude Code limits faster than before?
Effective capacity depends on model and context size: Opus drains the allowance several times faster than Sonnet, and long conversations resend context every turn. Anthropic has also adjusted limit levels over time — /usage reflects the current policy.
Do Claude Code weekly limits exist on every plan?
Pro and Max subscriptions have weekly caps in addition to the 5-hour window; Max plans also have a separate Opus cap. API-key (pay-per-token) usage has rate limits but no subscription-style weekly cap.
What's the best way to use less of my limit without working less?
Use smaller models for mechanical tasks, keep sessions short and contexts compact (/compact helps), and batch related questions. Model choice and context size dominate everything else.
Last updated 2026-07-07. Corrections welcome on GitHub.