Claude Code usage limits, and what resets when

The two windows

Anthropic's error reference describes a subscription plan's allowance as rolling, and Claude Code reads two windows out of it. The session limit runs on a five-hour window and the weekly limit runs on a week. Both are shared across Claude and Claude Code, so the same allowance covers your work in the terminal and your chat in the app.

Usage counts against both at the same time. Anthropic says a single burst of heavy activity, such as a large workflow fan-out, when many subagents work at once, can exhaust the weekly allowance before the session window resets. The two reset independently, so a fresh five-hour window does not give you a new week.

Limits that cover only one model family

More limits are scoped to a single family. Opus and Sonnet each have one, and Anthropic's error reference says each applies only to requests to that model family, so switching to a model outside the family with /model keeps you working. Anthropic's support pages add a third: on a Max plan there is also a separate weekly limit for Fable. That switch has a cost of its own. Anthropic notes each model keeps its own prompt cache, so the next request re-reads the whole conversation with no cache hits.

Team and Enterprise plans

Anthropic's cost documentation says a Team or Enterprise member's Claude Code usage draws on a per-seat allowance, resetting on the same rolling five-hour and weekly windows and shared with Claude chat and Cowork. The size depends on the seat tier. Anthropic's support pages add the exception worth knowing: an organization on a usage-based Enterprise plan has no per-seat limits at all, and pays by consumption at API rates.

What the limit message says

Claude Code prints these lines from Anthropic's error reference. The words after the dot are the reset time, and that is the part to read: it is the moment access comes back, not a countdown you have to guess at.

You've hit your session limit · resets 3:45pm

You've hit your weekly limit · resets Mon 12:00am

You've hit your Opus limit · resets 3:45pm

You've hit your Sonnet limit · resets 3:45pm

What each one means. The session and weekly lines cover every model and block further requests until the reset. The Opus and Sonnet lines cover only that family, so a switch with /model is a way through them.

Before the window runs out

Claude Code warns you first. Anthropic's error reference gives the wording as You've used 85% of your session limit · resets 3:45pm. To watch the figure continuously, Anthropic points at two places: the rate_limits fields on a custom status line, and the usage ring beside the model picker in the Desktop app, whose popover shows your plan usage.

The status-line fields are precise about what you can read. Anthropic's status-line reference lists rate_limits.five_hour.used_percentage and rate_limits.seven_day.used_percentage, each from 0 to 100, with a matching resets_at giving the window's reset as a Unix timestamp (seconds since 1 January 1970). It also says where they come from: the object only appears for claude.ai Pro and Max subscribers (or behind a Claude apps gateway that sets a spend limit for you), only after the first API response in the session, and each window can be absent on its own.

If the message does not say "usage limit"

Not every block is your plan. Anthropic's error reference separates three near neighbours. API Error: Server is temporarily limiting requests (not your usage limit) is a short-lived throttle on the API that Claude Code tells apart from your plan limit by the absence of the quota headers a real limit carries (the usage figures Anthropic sends alongside one), and it is retried automatically. API Error: Request rejected (429) is the rate limit configured for your API key, Amazon Bedrock project or Google Cloud project, and Anthropic's guidance is to check that provider's console for the active limits and request a higher tier there if you need one. And API Error: Repeated 529 Overloaded errors is a different thing again: the API is at capacity for everyone, which Anthropic's reference says is not your usage limit and does not count against your quota.

Two more come from a gateway rather than a plan. If you connect through a Claude apps gateway, a lowercase spend limit reached (daily; resets 2026-08-09 00:00 UTC) is a cap that gateway's operator set. A related line, spend limit unavailable, means the gateway could not read its own spend records and blocked the request as a precaution.

If you arrived here with a message already on your screen that is not one of these, the Claude Code error messages guide covers seventeen of them one by one, each with where its wording comes from.

What you can actually do

Four of these six options cost money, and one of them is easy to do by accident. The first two come from Anthropic's error reference and interactive-mode documentation, options three to five are the routes its support pages name for a Pro or Max user at the limit, and the sixth is ours, not a route Anthropic lists.

1. Wait for the reset

Access returns at the time the message names, without anyone raising anything. On a Team or Enterprise plan with a seat-based allowance the same is true, because the window is what ran out. Waiting costs nothing and needs nobody's permission.

2. Let Claude Code wait for you

Anthropic's interactive-mode documentation says that in an interactive session signed in with a claude.ai subscription, Claude Code can wait in the open session and continue the interrupted task shortly after the reset. The lines read Usage limit reached · limit resets 3:45pm and Continuing automatically at 3:45pm · esc to cancel. Anthropic says automatic continue is on by default from Claude Code 2.1.234, that it arms the wait at most twice in a row after repeated hits, and that you cancel it by pressing Esc at an empty prompt or by opening /rate-limit-options and choosing Don't continue automatically. Keep the session open: the wait does not restart if you exit.

3. Use usage credits

Anthropic's support article on usage credits says they let you keep working after an included limit by switching that usage to pay-as-you-go at standard API rates. They are separate from your subscription and appear as additional charges. Anthropic's error reference says /usage-credits buys additional usage on Pro and Max, and on Team and Enterprise it sends the request to your admin instead. Once your plan's included usage is spent, Anthropic says requests switch to credits and keep working until your monthly spend limit or your balance runs out.

4. Change plan

Anthropic's support page for Pro and Max suggests moving up a tier if you hit limits consistently: Max 5x if you are on Pro, Max 20x if you are on Max 5x. The command is /upgrade, which opens the upgrade page in your browser. Anthropic's documentation also notes /upgrade does not appear on Enterprise plans, where seat tier and organization settings decide the allowance instead.

5. Move to an API key

Anthropic lists switching to a Claude Console account and buying API usage credits as an option for intensive work. This is a different billing arrangement rather than a bigger allowance: API usage is billed per token at standard API rates, and Anthropic's rate-limits reference says an organization's API limits are set by usage tier, with a monthly spend cap per tier and a rate_limit_error once the cap is reached. Anthropic notes one trap on the subscription side: an ANTHROPIC_API_KEY in your environment routes requests through that key instead of your subscription, so a stray one can send you to the wrong bill. Its support pages also say the monthly API credits included with Max plans and with Team plans cannot be used in Claude Code.

6. More than one account

Anthropic ties the allowance to the plan (Subscription plans include a rolling usage allowance.), so as we read it a second subscription carries its own allowance. Anthropic's consumer terms draw the line that matters here: you may not share your account login information, API key or credentials with anyone else, or make your account available to anyone else. That is our reading of the terms, not a line from them: a second subscription you hold and sign into yourself is a separate account, and sharing one login between people is what the rule forbids.

Running out with more than one session open

The limits are per plan, not per session, so ten sessions in parallel spend one allowance ten times as fast. That is the case where a limit is most expensive to hit, because it arrives while several pieces of work are in flight.

Omniscio, the desktop app we make for running many Claude Code sessions at once, has its own answer to a worn-out allowance, and it belongs to one part of the app: managed accounts. That is an owner-side capability, not something every install has, so it is worth being plain about who gets it. A fresh install that has never added an account does not have it, and a sign-in that uses a single Claude account keeps using that one.

Where it is in play, a session that hits a usage limit waits rather than fails. Omniscio parks it in a visible waiting state, sets a timer, and picks the task back up on its own when the window opens. The session shows as waiting in the sidebar's Interrupted section with an orange dot, and while it waits you get no per-session inbox rows for it, because nothing about it needs you. Holding several logins goes further still: Omniscio can keep a pool of them, start each new session on the one with the most room left, and say so in the sidebar beside a count of the sessions waiting when every one has run out.

The page on sessions stuck in “needs you” covers the waiting behaviour in full, and the account pool page covers what holding several logins does. The features page lists the rest of the app, and the download page has the installers.

Sources

Every Anthropic message, meaning and step on this page rests on one of these, and the date beside each one is the day we read it. The two places where the page goes past them are marked as ours where they appear, and they are the sixth option and the Omniscio section.

• Claude Code error reference, for the session, weekly, Opus and Sonnet limit messages, the reset behaviour, the model-family carve-out, the 85% warning, the gateway spend cap, the 429, the 529, the temporary server throttle, and /usage and /usage-credits. Read October 22, 2026.
• Costs and usage, for the rolling allowance, the shared Claude and Claude Code usage, and the Team and Enterprise five-hour and weekly windows. Read October 22, 2026.
• Interactive mode, for waiting out a limit, the two lines it shows, the automatic continue and how to cancel it, and the re-arm limit. Read October 22, 2026.
• Status line, for the rate_limits fields, their units, and who they appear for. Read October 22, 2026.
• Commands, for /usage, /usage-credits, /rate-limit-options, /model and /upgrade. Read October 22, 2026.
• Desktop app, for the usage ring beside the model picker. Read October 22, 2026.
• Anthropic API rate limits, for usage tiers, the monthly spend cap and rate_limit_error. Read October 22, 2026.
• Using Claude Code with your Pro or Max plan, for the five-hour session limit and the weekly limit, the separate weekly limit for Fable on Max plans, and the options at the limit. Read October 22, 2026.
• Manage usage credits for paid Claude plans, for what usage credits are, that they bill at standard API rates, and the monthly spend limit. Read October 22, 2026.
• Use Claude Code with your Team or Enterprise plan, for seat-based allowances and the usage-based Enterprise exception. Read October 22, 2026.
• Anthropic consumer terms, for the account rule quoted above. Read October 22, 2026.

Anthropic moves both the limits and the wording. The error reference keeps notes such as "Before v2.1.234", the reset behaviour gained its automatic continue in that release, and the messages above are the current wording rather than a permanent one. Read the source again before you trust a copy of this page from a few months back.

Common questions

What are the Claude Code usage limits?

Anthropic describes a subscription plan's allowance as rolling, and Claude Code shows two windows from it: a five-hour session limit and a weekly limit. Those two are shared across all models, so switching models does not restore access. Opus, Sonnet and (on a Max plan) Fable each also carry their own cap.

How long is a Claude Code session limit?

The session limit runs on a five-hour window. Anthropic's support pages call it the five-hour session limit, and the message Claude Code prints names the exact reset time.

What does "You've hit your session limit" mean?

Your plan's rolling allowance has run out. Claude Code blocks further requests until the reset time the message shows, and the session and weekly limits are shared across all models, so switching models does not bring access back.

What is the difference between the session limit and the Opus or Sonnet limit?

The session and weekly limits cover every model. The Opus and Sonnet limits each apply only to requests to that model family, so switching to a model outside the family with /model keeps you working.

How do I check how much of my limit is left?

Run /usage to see your plan's limits and when each one resets. Claude Code can also warn you before a window runs out, with a line such as You've used 85% of your session limit. A custom status line can show the remaining allowance continuously.

What happens to a session that is mid-task when the limit hits?

In an interactive session signed in with a claude.ai subscription, Claude Code can wait in the open session and continue the interrupted task shortly after the reset. Anthropic says automatic continue is on by default from Claude Code 2.1.234 and can be cancelled.

Can I run Claude Code on an API key instead?

Yes. Anthropic lists switching to a Claude Console account and buying API usage as one of the options at the limit. API keys are billed per token at standard API rates rather than against a subscription allowance, and they carry their own rate limits by usage tier.

Does the weekly limit reset at the same time as the session limit?

No. They are separate windows, and Anthropic notes that usage counts against both at once, so a single burst of heavy activity can exhaust the weekly allowance before the session window resets.