Guides

Claude Code usage limits explained: 5-hour and weekly limits

On a Claude subscription, Claude Code counts your usage against two limits at once: a rolling five-hour window and a weekly limit. Both are shared with Claude on the web, desktop and phone. Anthropic publishes no message or token counts, only multiples: Max gives 5x or 20x Pro’s usage per five-hour session. Check what’s left with /usage or in Settings → Usage on claude.ai. At the limit, wait (Claude Code carries on by itself after the reset), turn on usage credits, upgrade, or switch to an API key.

By Dmitry Mishunin, who builds Meowdoro · October 30, 2026 · checked against Anthropic’s documentation and help center that day

• How the two limits work
• How much each plan gets
• Check what’s left
• When you hit a limit
• What uses it up faster
• Changes in 2025 and 2026
• Questions

How the two limits work

Most people watch only the 5-hour limit. That’s not enough: every request counts against both limits at the same time, and whichever runs out first stops you, sometimes the weekly one.

• The 5-hour limit resets on a rolling five-hour window. Claude Code used to call it the “session limit”; since version 2.1.312 (October 27, 2026) its messages say “5-hour limit”. Anthropic doesn’t say for Pro or Max whether the window starts with your first message. Only its pages for Claude for Government seats say so, with resets on the hour, so go by the reset time Claude Code shows you.
• The weekly limit applies on paid plans across all models and resets at a fixed weekly time assigned to your account, not on a calendar day. Settings → Usage shows when.
• One pool. Chats on the web, the desktop and mobile apps, and Claude Code all draw from the same limits. A long research chat in the morning leaves less for coding in the afternoon.
• Fable has its own weekly cap. On Max, Fable models can use up to half of your weekly limit, inside it rather than on top. On Pro, Fable isn’t part of the plan’s limits and runs on usage credits.

Claude Code’s error reference also lists an Opus limit and a Sonnet limit that apply only to that model family. Anthropic removed Opus-specific caps in November 2025 and no plan page lists family limits today, so you may never see these messages; if you do, switching model family helps, which isn’t true of the other two limits.

How much each plan gets

Anthropic describes usage only relative to other plans. How far that takes you depends on how long and complex your conversations are, the model, and the features you use, so there’s no fixed message count (Claude pricing).

The multiples are per five-hour session; Anthropic publishes no weekly multiple. Hour estimates you may find quoted, such as 40 to 80 hours of Sonnet a week on Pro, came from Anthropic’s comments to the press in July 2025, for models and limits that have changed since. Max plans also include monthly API credits, but those can’t be spent in Claude Code.

How to check what’s left

• /usage in Claude Code (/cost and /stats open the same screen) shows your plan usage bars, activity stats, and a breakdown of what’s eating your limits: parallel sessions, subagents, skills, plugins and each MCP server, as shares of the total. Press d or w for the last day or week. The breakdown counts only this machine, not other computers or claude.ai. Ignore the dollar figure in the Session block on a subscription: it’s an API-price estimate, not your bill.
• Settings → Usage on claude.ai shows progress bars for the five-hour and weekly limits and when each resets, for everything on your account.
• The status line. On Pro and Max, Claude Code passes both windows to your status line script as percentages with reset times, so you can keep them in view at all times. See how to show limits in the Claude Code status line.
• Warnings. As you get close, Claude Code warns you, for example “You’ve used 85% of your session limit · resets 3:45pm”, and since October 19, 2026 (version 2.1.304) it warns before resuming an old conversation that would use a noticeable share of your five-hour limit.

An older help article still points to /status for your remaining allocation. In current versions, /status shows the version, model, account and connectivity; the limits are in /usage.

What happens when you hit a limit

Claude Code stops sending requests and shows which limit you hit and when it resets, for example “You’ve hit your weekly limit · resets Mon 12:00am”. Where it’s rolling out, Claude Code may first finish the step it’s on (“Usage limit reached · wrapping up”); on Pro that happens at most once a week. /rate-limit-options lists your choices:

• Wait. In an interactive session signed in with a subscription, Claude Code waits and continues the task by itself after the reset (version 2.1.234 or later, on by default). A five-hour reset is a good length for a real break; a weekly one can be days away.
• Turn on usage credits (formerly “extra usage”) with /usage-credits. They’re billed at standard API rates, cover Claude Code and chat, don’t move your reset times, and you can cap them with a monthly limit. Team and Enterprise members without billing access send a request to their admin instead.
• Upgrade: Pro to Max 5x, Max 5x to Max 20x, or a Standard Team seat to Premium.
• Switch to an API key. With ANTHROPIC_API_KEY set, Claude Code uses the key instead of your subscription and bills per token, with no window to wait for. For scale, Anthropic puts the average for enterprise teams on the API at about $13 per developer per active day, and under $30 for 90% of users. Check that the key is unset when you want the subscription back.

Switching models doesn’t help with the five-hour or weekly limit: both cover every model.

What uses it up faster

Anthropic’s guidance comes down to one rule: “the more context Claude processes, the more tokens you use.” In practice (managing costs in Claude Code):

• Long conversations. Each message carries the whole conversation, so a one-line question in a session that’s been open all day costs like a big one: it’s like taking every book off the shelf to check one page. Use /clear between unrelated tasks; when to compact and when to clear covers the trade-offs. Note that /compact itself reads the whole conversation, so compacting a large context is a large request.
• Coming back from a break. The first message after a pause longer than the prompt cache’s lifetime reprocesses your full context. A fresh session after a long break is often cheaper than picking up a huge one.
• The model and its effort. Opus costs several times more per turn than Sonnet; higher effort and thinking add output tokens.
• Subagents, workflows and agent teams. Each sends its own requests on top of the main conversation. Anthropic puts agent teams at about 7x the tokens of a normal session when teammates run in plan mode, and says one burst of heavy activity can exhaust the weekly allowance before the five-hour window resets.
• Several sessions at once. They all draw from the same limits. We build Meowdoro this way, with separate Claude Code sessions for the app, the site and the videos, all on one plan and one window. Running several sessions is worth it; just count it in.
• Pasted files and logs. Anything you paste stays in context for the rest of the session. Let a hook or grep pick out the lines that matter.
• Always-loaded context. Keep CLAUDE.md under about 200 lines. MCP tool definitions now load only when used, but a CLI like gh is still leaner than an MCP server.
• Work while you’re away. A scheduled task or /loop sends the full context each time it fires, even when the session is idle.
• Fast mode bills usage credits directly, even with plan usage left.

Changes in 2025 and 2026

Some of the 2026 increases are worded as applying to Claude Code only, even though chat and Claude Code share one pool; Anthropic doesn’t explain how the two fit together.

Where Meowdoro fits

We built Meowdoro, a focus timer around the MacBook notch that shows your Claude Code and Codex sessions. One honest limitation: it doesn’t show Claude’s plan limits, because it would have to read your login from the Keychain, and it doesn’t. It does show how full each session’s context is, and its break card suggests /compact for a session that’s filling up, so the agent tidies up while you rest (how it shows agents). For the limits themselves, use the status line or /usage. And when you hit one, Claude Code waits for the reset by itself; that’s a good moment for a proper break. Try it and tell us what’s missing.

Questions

Does Claude Code have a weekly limit?

Yes, on every paid plan, on top of the five-hour limit. It covers all models and resets at a fixed weekly time assigned to your account, shown in Settings → Usage.

When does the Claude Code 5-hour limit reset?

Five hours into its window, at the time Claude Code prints in the limit message and in /usage. Anthropic doesn’t say for Pro and Max exactly when the window starts.

Are Claude Code limits shared with claude.ai?

Yes. Chats on the web, desktop and phone and Claude Code all draw from one pool on Pro, Max and Team.

How do I check my Claude Code usage?

Run /usage in Claude Code, open Settings → Usage on claude.ai, or show both windows in your status line.

Does switching to a cheaper model help when I hit the limit?

No. The five-hour and weekly limits cover all models, so another model hits the same wall. It helps only with an Opus or Sonnet family limit. A lighter model does make the next window last longer.

Can I use the API credits that come with Max in Claude Code?

No. Anthropic says the monthly API credits included with Max can’t be used in Claude Code or the Claude apps. Usage credits are a separate, pay-as-you-go option.

How do Claude Code’s limits compare with Codex’s?

Both have a five-hour and a weekly window, but OpenAI publishes rough message counts and Anthropic doesn’t. See Claude Code vs Codex usage limits.

Related guides

• Claude Code status line: show context, cost and limits

  Set it up, what your script gets, and examples to copy.

• Claude Code vs Codex usage limits

  Windows, plans, checking and running out, side by side.

• Claude Code /compact vs /clear vs a new session

  What survives /compact, and when a clean start is better.

Sources: Anthropic’s pricing page; the Claude Code documentation on costs, errors, commands, interactive mode and the changelog; help center articles on Claude Code with Pro or Max, the Max plan, the Pro plan, Fable models, usage credits and weekly limits in Claude Code; Anthropic’s posts on Opus 4.5, higher limits and the April 23 postmortem; TechCrunch on the 2025 weekly limits. Read on October 30, 2026.