Claude Code says you hit your weekly limit? When it resets, what counts, what to do

You are mid-task, the agent is on a roll, and the terminal prints You've hit your weekly limit. The forums will tell you it resets in 72 hours, or on Monday, or when the moon is right. Anthropic's own pages are clearer than that, and they are the only source I used here.

aliteq is published independently. Claude and Claude Code are made by Anthropic, which did not review this page. New to the tool? Start with Claude Code for vibe coders.

What are Claude Code's weekly and 5-hour limits?

They are two separate allowances on every paid Claude plan. The session limit resets every five hours. The weekly limit resets once a week, at a fixed time tied to your account. Your Claude Code work and your Claude chats draw from the same pool, and every request counts against both clocks at the same time.

Anthropic's pricing page puts it in one line: "Every plan has usage limits that reset on a rolling five-hour session window, and paid plans add weekly limits on top." The Pro, Max and Team help articles each say the weekly limit "applies across all models."

Three details from Anthropic's pages matter most:

• It's shared. The usage-limits article says use of "claude.ai, Claude Code, Claude Desktop" counts toward the same limit. A long chat in the browser eats into your coding week.
• Both clocks run at once. The Claude Code error docs warn that "a single burst of heavy activity, such as a large workflow fanout, can exhaust the weekly allowance before the session window resets."
• There is no fixed number. The pricing FAQ says "there's no fixed message count." Usage depends on conversation length, the model, the features you use and the effort level.

Anthropic also keeps a right to do more. All three plan pages say it "may limit your usage in other ways, such as weekly and monthly caps or model and feature usage, at our discretion."

When does the Claude Code weekly limit reset?

At a fixed day and time each week that Anthropic assigns to your account. It does not start counting when you first open Claude Code, and it does not follow your billing date. To find yours, open Settings > Usage on claude.ai. The limit message in Claude Code also prints the reset time.

The Pro, Max and Team articles use the same words: "Your reset day and time stay the same regardless of when you start using Claude or when your subscription begins, and you receive your full weekly allowance each cycle."

So the "72 hours" and "every Monday" claims you may have read are guesses. Your reset is whatever your account shows. The Claude Code docs print a sample message with a made-up time, You've hit your weekly limit · resets Mon 12:00am. Yours will name your own slot.

There is one exception. Anthropic sometimes gives eligible plans a free "limit reset," which sets the five-hour or the weekly limit back to full. The help article says the "Reset for free" button is not in Claude Code in the terminal or IDE. Use the web or Claude Desktop, and Claude Code is back to full too, since the limits are shared.

How big is the weekly limit on Pro, Max and Team?

Anthropic does not publish it as tokens, messages or hours. It publishes multiples of the Pro plan's per-session allowance: Max 5x gets five times, Max 20x gets 20 times, a Team Standard seat 1.25 times and a Team Premium seat 6.25 times. No page states the weekly multiple between plans.

Two notes. Team limits are per member: if one teammate runs out, the others keep working. And the two Team numbers agree with the pricing page, which says Premium seats get "5x more usage than standard seats." Five times 1.25 is 6.25.

If you use Fable models, there is a second rule inside the week. On Max and Team Premium, the Fable help article says you can spend up to 50% of your weekly limit on Fable, and Fable uses it "faster than other Claude models." On Pro and Team Standard, Fable runs only on paid usage credits.

Did Anthropic cut the weekly limit?

Compared with the summer, yes. Compared with before May, no. Anthropic's promotion page says Claude Code's weekly limit was 50% higher from 13 May to 13 September 2026. From 14 September it is 25% higher than before the promotion. By my arithmetic, today's limit is 16.7% below the promotion level.

The page's own words: "starting September 14, 2026, weekly limits in Claude Code are 25% higher than they were before the promotion for Pro, Max, Team, and seat-based Enterprise plans." It also says "5-hour usage limits were not affected by this promotion," and that the boost applied to Claude Code only, not to chat or Cowork.

That explains the "17% cut" you may have seen. If your habits were built during the summer, your week now ends sooner. It does not mean the limit is lower than it was in the spring.

People did notice. When the promotion page was posted to Hacker News on 18 August, the thread drew 267 comments. One commenter wrote: "I'm regularly hitting 90-100% on my $200/mo sub." The same commenter added: "This constant uncertainty is really annoying." Another said they hit weekly limits "two to three days in." These are individual users describing their own weeks, not measurements.

How do I check my Claude Code usage?

Run /usage inside Claude Code. On a Pro, Max, Team or Enterprise plan it shows your plan usage bars, activity stats and a breakdown of what used your allowance. On claude.ai, Settings > Usage shows the current-session bar, the weekly bar and when each resets. Either works mid-task.

From the Claude Code docs, what /usage gives a subscriber:

• Plan bars. How much of the session and weekly limits you have used. /cost and /stats open the same screen.
• Attribution. Recent usage split across skills, subagents, plugins and individual MCP servers, each as a percentage.
• Behavior flags. A flag appears when something like long context or cache misses makes up 10% or more of recent usage.
• A day or week view. Press d or w. The breakdown comes from session history on that machine only, so work on another computer or on claude.ai is not in it.

Ignore the dollar figure at the top. The docs say Pro and Max subscribers have usage included, "so the session cost figure isn't relevant for billing purposes." One more naming quirk: the Pro and Max help article says to monitor your allocation "using the /status command," but the Claude Code commands page describes /status as version, model, account and connectivity. /usage is the one that lists plan limits.

Want a live readout? The status-line docs expose a rate_limits object with a five_hour and a seven_day window, each with used_percentage and a resets_at time. It is available to Pro and Max subscribers after the first response of a session. Claude Code also warns you before a window runs out, with a line like You've used 85% of your session limit · resets 3:45pm.

What happens when you hit the limit?

Claude Code stops sending requests until the reset time in the message. Since version 2.1.234, an interactive session signed in with your subscription waits and continues the interrupted task by itself after the reset, as long as you leave it open. Nothing moves you onto paid usage without your say-so.

The docs list four messages, and they are not the same problem:

The auto-wait shows two lines, Usage limit reached · limit resets 3:45pm and Continuing automatically at 3:45pm · esc to cancel. At the reset, Claude Code prompts Claude to pick up where it stopped. The docs say it "doesn't resend your last message." It still stops for permission prompts, and it re-arms the wait at most twice in a row. Press Esc, or open /rate-limit-options, to cancel.

In the Claude Desktop app's Code tab, the session-limit card has an "Auto-continue when limits reset" box. The docs note that "the weekly-limit card doesn't." A weekly wait can be days, so plan for that.

On consent, the Pro and Max help article is explicit: "All transitions to API credit usage require explicit user consent." Decline the offer and you stay inside your plan.

How do I keep working after the weekly limit?

Four ways: wait for the reset, turn on usage credits, sign in with an API key, or move to a bigger plan. Only waiting is free. Usage credits and API keys both bill at standard API rates on top of your subscription. A model switch helps only with the Opus or Sonnet message, not the weekly one.

Usage credits. These keep you on your subscription login. The help article says you continue "at standard API rates," that the charges appear on your bill separately from the plan, and that there is a daily redemption limit of $2,000. You can disable them any time in Settings > Usage.

API key. Anthropic's Claude Code guide says an API key is "Pay-as-you-go, billed per token" with "No hard stop; the account is charged for what it uses." That is a feature in a crunch and a risk if you forget. One trap: if an ANTHROPIC_API_KEY environment variable is set on your machine, the Pro and Max help article says Claude Code uses it "instead of your Claude subscription," which means API charges even when you meant to stay on the plan. To stop the API-credit offer appearing at all, the article says to run claude logout, then claude login with only your plan credentials.

What does the per-token route cost? Anthropic's API price list, read 4 October 2026, per million output tokens: Opus 5.5 $20, Sonnet 5.5 $10, Haiku 4.5 $5. Input is $4, $2 and $1 per million. A token is roughly a short word fragment, and a long agent session moves a lot of them.

How do I make the weekly limit last longer?

Use a cheaper model for routine work and keep each conversation short. Anthropic's Claude Code guide says Opus "uses meaningfully more of your quota" and calls Opus on routine work "the fastest way to drain a daily limit." Clearing the chat between tasks helps, because every turn resends the whole conversation so far.

The guide lists five habits. Paraphrased:

• Clear between tasks. Run /clear when you switch jobs. Use /compact mid-task to shrink the history into a summary. Both are about the context window, the text the model rereads on every turn.
• Match the model to the job. Sonnet for most coding, Opus when you are stuck or the change is wide, Haiku for renames and boilerplate. /model opusplan plans with Opus and executes with Sonnet.
• Point at files instead of pasting them. Name the path. The guide notes that @ pulls in the whole file.
• Keep CLAUDE.md lean. It is sent on every turn. The guide suggests staying under roughly 200 lines.
• Ask for a plan before big changes. A wrong 400-line diff costs far more than a short plan.

If model choice feels like guesswork, which AI model for which job walks through it in plain terms. For how to give the agent less but better context, see context engineering for vibe coders. And if the /usage breakdown blames one MCP server or a chatty hook, Claude Code hooks, explained shows where those run.