[{"data":1,"prerenderedAt":50},["ShallowReactive",2],{"blog-claude-code-usage-limits":3,"blog-related-claude-code-usage-limits":33},{"slug":4,"title":5,"description":6,"category":7,"categoryId":8,"keyword":9,"date":10,"minutes":11,"body":12,"html":13,"headings":14},"claude-code-usage-limits","Claude Code usage limits: the 5-hour window and weekly cap","How Claude Code's 5-hour window and weekly limit work, what makes you hit them, how to check usage, and what to do the moment Claude Code stops.","Usage & limits","usage-limits","claude code usage limits","2026-11-17",5,"On a Claude Pro or Max plan, Claude Code runs against two limits: a rolling 5-hour session window and a weekly limit. They are shared with Claude chat on claude.ai, so a long conversation there counts too. When you hit one, Claude Code tells you when it resets, and in recent versions it can wait and continue the task by itself.\n\n## How do the 5-hour and weekly limits work?\n\nAnthropic's help center describes it this way for Pro and Max: there is a five-hour session limit and a weekly limit, and \"these limits are shared across Claude and Claude Code\". Some tiers differ (for example, a separate weekly limit for certain models), and the exact amounts change, so this article gives no numbers. Your plan's page on claude.com is the source of truth.\n\nThe practical model:\n\n- **The 5-hour window** is the one you feel day to day. It is a usage allowance that refills on a rolling basis, not a count of messages.\n- **The weekly limit** is the backstop. Heavy days feed it; a hard Tuesday can leave less for Friday.\n- **Usage is tokens, not prompts.** One short prompt that makes Claude read thirty files can cost more than twenty chat messages.\n\nOn Team and Enterprise, each seat has an allowance on the same kind of rolling five-hour and weekly windows, shared with chat. Admins can add usage credits and spend limits on top.\n\n## How do I check how much I have used?\n\nRun `\u002Fusage` in Claude Code. On a subscription plan it shows plan usage bars, activity stats and a breakdown of what counted against your limits: skills, subagents, plugins and individual MCP servers, each as a percentage, plus recommendations. Press `d` or `w` to switch between the last 24 hours and the last 7 days.\n\nThree caveats from the docs:\n\n- The figures are computed from session history on this machine, so usage on other devices or on claude.ai is not in the breakdown, even though it counts against your limits.\n- The Session block at the top of `\u002Fusage` (a dollar figure) is meant for API users. As a subscriber, ignore it for billing purposes.\n- `\u002Fstatus` shows which account and credential Claude Code is using.\n\nIf `\u002Fusage` shows a message like \"rate limited until 3:14am\", that means Claude Code cannot fetch your usage numbers until then, not that your work is blocked.\n\n## What makes me hit the limit faster?\n\nThe docs list the causes of a session using more than your activity suggests:\n\n- **Long context.** The full conversation is sent with every request, and every tool call is another request. See [Claude Code's context window](\u002Fblog\u002Fclaude-code-context-window).\n- **Cache misses.** Your first message after a break longer than the cache lifetime reprocesses everything.\n- **Subagents and agent teams.** Each one makes its own requests on top of the main conversation. Agent teams, where enabled, multiply this roughly by the number of teammates.\n- **Scheduled tasks and loops.** A `\u002Floop` or other session-scoped schedule fires even while you are away, with your full context each time.\n- **Parallel sessions.** Five sessions is five times the burn on one plan. See [running several sessions at once](\u002Fblog\u002Frun-multiple-claude-code-sessions).\n- **Extended thinking and high effort.** Thinking tokens are billed as output. Lower effort with `\u002Feffort` for simple work.\n- **The model you pick.** Bigger models use more of your allowance for the same job.\n\n`\u002Fusage` flags behaviors behind 10% or more of your recent usage, which is the fastest way to find your own culprit.\n\n## What happens when I hit the limit?\n\nYou see a message such as \"You've hit your session limit\" or \"You've hit your weekly limit\", with the reset time. Switching models with `\u002Fmodel` does not help for these, because the window is shared across models. There are separate messages when a limit applies to a single model family; in that case a model outside that family keeps working.\n\nClaude Code v2.1.234 and later add a better option. In an interactive session signed in with a claude.ai subscription, it waits and continues on its own after the reset:\n\n```text\nUsage limit reached · limit resets 3:45pm\nContinuing automatically at 3:45pm · Esc to cancel\n```\n\nKeep the session open. At the reset it prompts Claude to pick the task up where it stopped (it does not resend your last message). If your computer slept for more than about 30 minutes across the reset, it asks you to press Enter instead. Permissions still apply, so a task can stop on a prompt while you are away. After three automatic continuations it stops and points you to `\u002Frate-limit-options`, where you can choose what to do next or pick \"Don't continue automatically\". An admin setting, `autoContinueAtUsageLimit`, controls this for a team.\n\nOther ways out, in order of effort:\n\n1. **Wait for the reset** and do something that does not need Claude: review the diff, write the next prompt.\n2. **Add usage credits** with `\u002Fusage-credits`. It opens your usage settings on claude.ai, where you can turn credits on and set a monthly spend limit. They bill at API rates, separate from the plan.\n3. **Use another account** that is genuinely yours (a team seat, say). See [using several Claude accounts](\u002Fblog\u002Fclaude-code-multiple-accounts).\n4. **Change plan** if it keeps happening. See [Claude Pro or Max for Claude Code](\u002Fblog\u002Fclaude-code-pro-vs-max).\n\nOne trap: if `ANTHROPIC_API_KEY` is set in your environment, Claude Code uses the key after you approve it and bills the API instead of your plan. `unset ANTHROPIC_API_KEY` and check `\u002Fstatus` if that is not what you want.\n\n## How do I make the window last longer?\n\n- `\u002Fclear` between unrelated tasks; `\u002Fcompact \u003Cfocus>` in long ones.\n- Match the model and effort to the job. Routine edits do not need the heaviest setting.\n- Give specific prompts and verification targets so there are fewer rework loops.\n- Use plan mode before big changes, so a wrong direction costs a plan, not an implementation.\n- Send verbose work (test runs, log digging, docs lookups) to a subagent on a smaller model.\n- Cap what runs unattended: loops, routines and background agents burn the same pool.\n- Press Esc early when Claude heads the wrong way. `\u002Frewind` restores code and conversation.\n\nEach of these takes seconds. A long session left to grow without them is what drains the window.\n\n## In Relais\n\nRelais runs your own Claude Code login, so everything it starts counts toward the same 5-hour window and weekly limit, exactly as in a terminal. It shows your window under your account in the sidebar with its reset time. A session stopped by the limit carries on by itself about a minute after the reset, with the computer kept awake; with a second account or organization added, it carries on there at once. Agents also stop starting new work past a share of your plan (85% by default, adjustable). More in the [FAQ](\u002Ffaq) and on [features](\u002Ffeatures).","\u003Cp>On a Claude Pro or Max plan, Claude Code runs against two limits: a rolling 5-hour session window and a weekly limit. They are shared with Claude chat on claude.ai, so a long conversation there counts too. When you hit one, Claude Code tells you when it resets, and in recent versions it can wait and continue the task by itself.\u003C\u002Fp>\n\u003Ch2 id=\"how-do-the-5-hour-and-weekly-limits-work\">\u003Ca class=\"anchor\" href=\"#how-do-the-5-hour-and-weekly-limits-work\" aria-label=\"Link to this section\">#\u003C\u002Fa>How do the 5-hour and weekly limits work?\u003C\u002Fh2>\n\u003Cp>Anthropic&#39;s help center describes it this way for Pro and Max: there is a five-hour session limit and a weekly limit, and &quot;these limits are shared across Claude and Claude Code&quot;. Some tiers differ (for example, a separate weekly limit for certain models), and the exact amounts change, so this article gives no numbers. Your plan&#39;s page on claude.com is the source of truth.\u003C\u002Fp>\n\u003Cp>The practical model:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>The 5-hour window\u003C\u002Fstrong> is the one you feel day to day. It is a usage allowance that refills on a rolling basis, not a count of messages.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>The weekly limit\u003C\u002Fstrong> is the backstop. Heavy days feed it; a hard Tuesday can leave less for Friday.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Usage is tokens, not prompts.\u003C\u002Fstrong> One short prompt that makes Claude read thirty files can cost more than twenty chat messages.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>On Team and Enterprise, each seat has an allowance on the same kind of rolling five-hour and weekly windows, shared with chat. Admins can add usage credits and spend limits on top.\u003C\u002Fp>\n\u003Ch2 id=\"how-do-i-check-how-much-i-have-used\">\u003Ca class=\"anchor\" href=\"#how-do-i-check-how-much-i-have-used\" aria-label=\"Link to this section\">#\u003C\u002Fa>How do I check how much I have used?\u003C\u002Fh2>\n\u003Cp>Run \u003Ccode>\u002Fusage\u003C\u002Fcode> in Claude Code. On a subscription plan it shows plan usage bars, activity stats and a breakdown of what counted against your limits: skills, subagents, plugins and individual MCP servers, each as a percentage, plus recommendations. Press \u003Ccode>d\u003C\u002Fcode> or \u003Ccode>w\u003C\u002Fcode> to switch between the last 24 hours and the last 7 days.\u003C\u002Fp>\n\u003Cp>Three caveats from the docs:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>The figures are computed from session history on this machine, so usage on other devices or on claude.ai is not in the breakdown, even though it counts against your limits.\u003C\u002Fli>\n\u003Cli>The Session block at the top of \u003Ccode>\u002Fusage\u003C\u002Fcode> (a dollar figure) is meant for API users. As a subscriber, ignore it for billing purposes.\u003C\u002Fli>\n\u003Cli>\u003Ccode>\u002Fstatus\u003C\u002Fcode> shows which account and credential Claude Code is using.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>If \u003Ccode>\u002Fusage\u003C\u002Fcode> shows a message like &quot;rate limited until 3:14am&quot;, that means Claude Code cannot fetch your usage numbers until then, not that your work is blocked.\u003C\u002Fp>\n\u003Ch2 id=\"what-makes-me-hit-the-limit-faster\">\u003Ca class=\"anchor\" href=\"#what-makes-me-hit-the-limit-faster\" aria-label=\"Link to this section\">#\u003C\u002Fa>What makes me hit the limit faster?\u003C\u002Fh2>\n\u003Cp>The docs list the causes of a session using more than your activity suggests:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>Long context.\u003C\u002Fstrong> The full conversation is sent with every request, and every tool call is another request. See \u003Ca href=\"\u002Fblog\u002Fclaude-code-context-window\">Claude Code&#39;s context window\u003C\u002Fa>.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Cache misses.\u003C\u002Fstrong> Your first message after a break longer than the cache lifetime reprocesses everything.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Subagents and agent teams.\u003C\u002Fstrong> Each one makes its own requests on top of the main conversation. Agent teams, where enabled, multiply this roughly by the number of teammates.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Scheduled tasks and loops.\u003C\u002Fstrong> A \u003Ccode>\u002Floop\u003C\u002Fcode> or other session-scoped schedule fires even while you are away, with your full context each time.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Parallel sessions.\u003C\u002Fstrong> Five sessions is five times the burn on one plan. See \u003Ca href=\"\u002Fblog\u002Frun-multiple-claude-code-sessions\">running several sessions at once\u003C\u002Fa>.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Extended thinking and high effort.\u003C\u002Fstrong> Thinking tokens are billed as output. Lower effort with \u003Ccode>\u002Feffort\u003C\u002Fcode> for simple work.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>The model you pick.\u003C\u002Fstrong> Bigger models use more of your allowance for the same job.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Ccode>\u002Fusage\u003C\u002Fcode> flags behaviors behind 10% or more of your recent usage, which is the fastest way to find your own culprit.\u003C\u002Fp>\n\u003Ch2 id=\"what-happens-when-i-hit-the-limit\">\u003Ca class=\"anchor\" href=\"#what-happens-when-i-hit-the-limit\" aria-label=\"Link to this section\">#\u003C\u002Fa>What happens when I hit the limit?\u003C\u002Fh2>\n\u003Cp>You see a message such as &quot;You&#39;ve hit your session limit&quot; or &quot;You&#39;ve hit your weekly limit&quot;, with the reset time. Switching models with \u003Ccode>\u002Fmodel\u003C\u002Fcode> does not help for these, because the window is shared across models. There are separate messages when a limit applies to a single model family; in that case a model outside that family keeps working.\u003C\u002Fp>\n\u003Cp>Claude Code v2.1.234 and later add a better option. In an interactive session signed in with a claude.ai subscription, it waits and continues on its own after the reset:\u003C\u002Fp>\n\u003Cpre class=\"code\" data-lang=\"text\">\u003Ccode>Usage limit reached · limit resets 3:45pm\nContinuing automatically at 3:45pm · Esc to cancel\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>Keep the session open. At the reset it prompts Claude to pick the task up where it stopped (it does not resend your last message). If your computer slept for more than about 30 minutes across the reset, it asks you to press Enter instead. Permissions still apply, so a task can stop on a prompt while you are away. After three automatic continuations it stops and points you to \u003Ccode>\u002Frate-limit-options\u003C\u002Fcode>, where you can choose what to do next or pick &quot;Don&#39;t continue automatically&quot;. An admin setting, \u003Ccode>autoContinueAtUsageLimit\u003C\u002Fcode>, controls this for a team.\u003C\u002Fp>\n\u003Cp>Other ways out, in order of effort:\u003C\u002Fp>\n\u003Col>\n\u003Cli>\u003Cstrong>Wait for the reset\u003C\u002Fstrong> and do something that does not need Claude: review the diff, write the next prompt.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Add usage credits\u003C\u002Fstrong> with \u003Ccode>\u002Fusage-credits\u003C\u002Fcode>. It opens your usage settings on claude.ai, where you can turn credits on and set a monthly spend limit. They bill at API rates, separate from the plan.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Use another account\u003C\u002Fstrong> that is genuinely yours (a team seat, say). See \u003Ca href=\"\u002Fblog\u002Fclaude-code-multiple-accounts\">using several Claude accounts\u003C\u002Fa>.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Change plan\u003C\u002Fstrong> if it keeps happening. See \u003Ca href=\"\u002Fblog\u002Fclaude-code-pro-vs-max\">Claude Pro or Max for Claude Code\u003C\u002Fa>.\u003C\u002Fli>\n\u003C\u002Fol>\n\u003Cp>One trap: if \u003Ccode>ANTHROPIC_API_KEY\u003C\u002Fcode> is set in your environment, Claude Code uses the key after you approve it and bills the API instead of your plan. \u003Ccode>unset ANTHROPIC_API_KEY\u003C\u002Fcode> and check \u003Ccode>\u002Fstatus\u003C\u002Fcode> if that is not what you want.\u003C\u002Fp>\n\u003Ch2 id=\"how-do-i-make-the-window-last-longer\">\u003Ca class=\"anchor\" href=\"#how-do-i-make-the-window-last-longer\" aria-label=\"Link to this section\">#\u003C\u002Fa>How do I make the window last longer?\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Ccode>\u002Fclear\u003C\u002Fcode> between unrelated tasks; \u003Ccode>\u002Fcompact &lt;focus&gt;\u003C\u002Fcode> in long ones.\u003C\u002Fli>\n\u003Cli>Match the model and effort to the job. Routine edits do not need the heaviest setting.\u003C\u002Fli>\n\u003Cli>Give specific prompts and verification targets so there are fewer rework loops.\u003C\u002Fli>\n\u003Cli>Use plan mode before big changes, so a wrong direction costs a plan, not an implementation.\u003C\u002Fli>\n\u003Cli>Send verbose work (test runs, log digging, docs lookups) to a subagent on a smaller model.\u003C\u002Fli>\n\u003Cli>Cap what runs unattended: loops, routines and background agents burn the same pool.\u003C\u002Fli>\n\u003Cli>Press Esc early when Claude heads the wrong way. \u003Ccode>\u002Frewind\u003C\u002Fcode> restores code and conversation.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>Each of these takes seconds. A long session left to grow without them is what drains the window.\u003C\u002Fp>\n\u003Ch2 id=\"in-relais\">\u003Ca class=\"anchor\" href=\"#in-relais\" aria-label=\"Link to this section\">#\u003C\u002Fa>In Relais\u003C\u002Fh2>\n\u003Cp>Relais runs your own Claude Code login, so everything it starts counts toward the same 5-hour window and weekly limit, exactly as in a terminal. It shows your window under your account in the sidebar with its reset time. A session stopped by the limit carries on by itself about a minute after the reset, with the computer kept awake; with a second account or organization added, it carries on there at once. Agents also stop starting new work past a share of your plan (85% by default, adjustable). More in the \u003Ca href=\"\u002Ffaq\">FAQ\u003C\u002Fa> and on \u003Ca href=\"\u002Ffeatures\">features\u003C\u002Fa>.\u003C\u002Fp>\n",[15,18,21,24,27,30],{"id":16,"text":17},"how-do-the-5-hour-and-weekly-limits-work","How do the 5-hour and weekly limits work?",{"id":19,"text":20},"how-do-i-check-how-much-i-have-used","How do I check how much I have used?",{"id":22,"text":23},"what-makes-me-hit-the-limit-faster","What makes me hit the limit faster?",{"id":25,"text":26},"what-happens-when-i-hit-the-limit","What happens when I hit the limit?",{"id":28,"text":29},"how-do-i-make-the-window-last-longer","How do I make the window last longer?",{"id":31,"text":32},"in-relais","In Relais",[34,39,44],{"slug":35,"title":36,"description":37,"category":7,"categoryId":8,"keyword":38,"date":10,"minutes":11},"claude-code-context-window","Claude Code's context window: why it fills up and what to do","Why Claude Code's context window fills up, what loads before you type, and the commands (\u002Fcontext, \u002Fcompact, \u002Fclear) that stop you running out mid-task.","claude code context window",{"slug":40,"title":41,"description":42,"category":7,"categoryId":8,"keyword":43,"date":10,"minutes":11},"claude-code-pro-vs-max","Claude Pro or Max for Claude Code: which plan fits you?","Claude Pro or Max for Claude Code? How to decide from your own usage: how often you hit the limit, how many sessions you run, and what else you could do.","claude pro vs max claude code",{"slug":45,"title":46,"description":47,"category":7,"categoryId":8,"keyword":48,"date":10,"minutes":49},"claude-code-multiple-accounts","How to use several Claude accounts with Claude Code","Keep a personal plan and a work seat signed in to Claude Code at once with CLAUDE_CONFIG_DIR, switch cleanly, and avoid the API key trap. Commands included.","claude code multiple accounts",4,1794931098788]