What each Claude Code limit message means, and who can act on it

When a developer is stopped by Claude Code, the message on their screen names a specific limit, and the right response depends on which one. Anthropic's documentation on managing costs, read on 4 October 2026, has a section written for administrators called When a developer asks about a limit. It says developers usually bring limit questions to their administrator, and it lists five situations that, in Anthropic's words, mean different things [1]. This page takes each one in turn, gives the exact wording Anthropic's error reference shows, and says who can act. It is part of the guide to Claude Code costs for teams.

Four terms recur. A token is a piece of text the model reads or writes, and every limit on this page is measured either in tokens or in the dollars that tokens cost. On a Claude for Teams or Enterprise plan each member has a seat, a paid place on the plan, and each seat has an allowance: an amount of usage included in the seat's price, which Anthropic says resets on a rolling five-hour window and a weekly window and is shared with Claude chat and Cowork [1]. Usage credits are paid usage a member draws on after the allowance runs out, once an administrator has turned them on. Managed settings are settings an organisation deploys to every developer's machine, which a developer's own settings cannot override.

The five situations in one table

Message on the developer's screen	What it is	Who acts	Action
You've hit your session limit · resets 3:45pm, or You've hit your weekly limit · resets Mon 12:00am	A seat allowance window on a subscription plan has run out; shared across all models [1]	Developer, or Claude Code itself	Wait for the reset shown; run /usage-credits if usage credits are on; on version 2.1.234 or later, let Claude Code wait and continue automatically [1]
You've hit your Opus limit, or You've hit your Sonnet limit, each with a reset time	The allowance for one model family has run out [2]	Developer	Run /model and switch to a model outside that family [1]
You've hit your individual spend limit, You've hit your org's monthly spend limit, or You've hit your team's shared budget	The request would be billed to usage credits, and those credits have reached a spend limit the administrator set [1]	Administrator	Raise the limit the message names at Organization settings > Usage; the developer can wait instead when the message also names a plan reset time [1]
spend limit reached (daily; resets 2026-08-09 00:00 UTC), in lowercase	A spend cap set by the operator of a self-hosted Claude apps gateway [2]	Gateway operator	Wait for the period to reset, or the operator raises the cap [1]
A context or auto-compact warning	A different thing from a usage limit: the conversation is close to the auto-compact window [1]	Developer	Follow Anthropic's advice on reducing token usage; the sister guide covers it [1]

Anthropic adds a sixth entry to the same list that produces no message at all: unexpectedly high spend on an API or cloud-provider plan, which an administrator notices on a bill or a dashboard. It is covered below with the others.

Session and weekly limits: a seat allowance window

The first message reads You've hit your session limit, with a reset time such as resets 3:45pm, or You've hit your weekly limit, with a reset such as resets Mon 12:00am [2]. Anthropic's documentation says this is a seat-based usage window on a subscription plan, shared across all models, so the developer cannot restore access by switching models with /model [1]. The message shows when the window resets, and Claude Code blocks further requests until that time [2]. Anthropic's cost page describes the allowance as resetting on a rolling five-hour window and a weekly window, and its error reference names the two messages as the session limit and the weekly limit; by the plain reading the session limit is the five-hour window, and Anthropic's pages describe the two windows and the two messages without pairing them in one sentence [1][2].

One detail matters to a manager who sees a weekly limit reached early in the week. Anthropic's error reference says usage counts against the session and weekly allowances at the same time, and that a single burst of heavy activity, such as a large workflow fanout, can exhaust the weekly allowance before the session window resets [2]. A workflow fanout is a run in which Claude starts many copies of itself at once, each doing its own work. So a weekly limit reached early in the week can come from one afternoon of heavy automation.

Before a window runs out, Claude Code can warn the developer that most of it has been used, with a message such as You've used 85% of your session limit · resets 3:45pm [2]. Anthropic's documentation says the developer can do two things while a window is exhausted: run /usage-credits to request usage beyond the allowance, if usage credits are turned on, or, on Claude Code version 2.1.234 or later, let Claude Code wait in the open session and continue the interrupted task after the reset [1]. The wait is described in its own section below. The page on spend limits for an organisation, a group and a person explains what /usage-credits does for each role.

Opus and Sonnet limits: one model family

The second pair of messages reads You've hit your Opus limit or You've hit your Sonnet limit, each with a reset time [2]. Anthropic's documentation says these apply only to requests to that model family, and that after one of them, switching to a model outside that family with /model does keep the developer working [1]. So a developer who is stopped on Opus can continue on Sonnet until the Opus window resets.

The switch has a cost that an administrator should know about. Anthropic's error reference says each model has its own prompt cache, so the next request re-reads the whole conversation with no cache hits [2]. The prompt cache is a store of request text the model service has already processed; while the start of a request matches the store, that part is read at a lower price. A long conversation switched to a new model is processed in full once, and the sister guide's page on switching model or effort mid-task gives the detail.

Spend limits: usage credits reached a cap you set

The third group of messages names a spend limit: You've hit your individual spend limit · ask your admin for a higher limit, You've hit your org's monthly spend limit · visit claude.ai/admin-settings/usage to raise it, or You've hit your team's shared budget · ask your admin to raise it at claude.ai/admin-settings/usage [2]. Anthropic's documentation says these mean the developer's request would be billed to usage credits, and those credits have reached a spend limit the administrator set [1]. The text after the dot says how to get that limit increased and varies with the plan and whether the developer manages billing [2].

The action belongs to the administrator. To let the developer continue, Anthropic says, go to Organization settings > Usage and increase the limit the message names; when the message also names a plan reset time, the developer can instead wait until then [1]. A developer without billing access can run /usage-credits, which sends the request to the organisation's administrators [2]. Three details from Anthropic's error reference help an administrator read the message. A team's shared budget is a pooled budget assigned to a group the developer belongs to, and the message does not name the group [2]. On organisations with usage-based billing, the message says usage limit in place of spend limit, as in You've hit your individual usage limit [2]. And the wording has changed with versions: before Claude Code version 2.1.239 the message did not name the plan window's reset time, and before version 2.1.268 a group's pooled budget produced the individual spend limit message instead of team's shared budget [2]. A developer on an older version may report the wrong one of these.

The error reference also lists a Pro or Max message, You've hit your monthly spend limit · raise it at claude.ai/settings/usage, which an individual raises in their own settings, and a channel's monthly spend limit, which is the budget of the one Slack channel a session runs in [2]. Both are raised outside the Teams or Enterprise administrator's organisation limit: the first by the individual in their own settings, the second by an org owner or channel manager in the channel's Claude settings [2].

A gateway spend limit: lowercase, from your own gateway

The fourth situation is also a spend limit, set in a different place. Anthropic's error reference says that a developer who connects through a Claude apps gateway, a server the organisation runs itself between Claude Code and Anthropic, and who sees a lowercase spend limit reached, has passed a spend cap the gateway operator set [2]. The gateway blocks the developer's requests until the named period resets or the operator raises the cap [1]. The message names the cap's period and reset time, for example spend limit reached (daily; resets 2026-08-09 00:00 UTC), and when the operator configured a blocked message, their own instructions follow it [2]. Before Claude Code version 2.1.225 the message read only spend limit reached, and a gateway on an older version still sends that shorter form [2]. The gateway marks each blocked response so that Claude Code shows the message without retrying [2]. The capital letter is the quickest way to tell the two apart: Anthropic's own messages begin You've hit your, and the gateway's begins in lowercase. The caps themselves are explained on the spend limits page.

A context warning is a different thing

The fifth situation produces a warning about context or auto-compaction, and Anthropic's documentation says plainly that this is a different thing from a usage limit [1]. The context window is everything the model reads in one request: Anthropic's instructions, the project's notes, the whole conversation so far, and every file opened. Compaction is what Claude Code does when that window is close to full: it summarises older history to free space. Anthropic's documentation says the warning means the conversation has grown close to the session's auto-compact window, the threshold where Claude Code summarises older history, and its advice to the administrator is to direct the developer to its guidance on reducing token usage [1]. The billing screens stay as they are. The sister guide's page on clear vs compact vs rewind is the developer's page for this.

Unexpectedly high spend on an API or cloud plan

Anthropic's list ends with a situation that arrives as a number instead of a message: unexpectedly high spend on an API or cloud-provider plan. Anthropic's documentation says this usually traces back to long sessions that were never cleared or to Opus left as the default model, and that the highest-impact habits to share are clearing between unrelated tasks and matching the model to the job [1]. Both are developer habits, and the page on what Claude Code costs per developer sets out everything that moves the figure. A message that reads Request rejected (429) is a rate limit, a cap on how much work the organisation may send per minute, and the page on rate limits per user by team size covers it.

Waiting for a reset, and the setting that controls it

For the session, weekly, Opus and Sonnet limits, Claude Code can wait on the developer's behalf. Anthropic's interactive mode documentation, read on 4 October 2026, says that when a claude.ai usage limit stops Claude mid-task, Claude Code waits in the open session and continues the task on its own after the limit resets; automatic continue is on by default in interactive sessions signed in with a claude.ai subscription, and it requires Claude Code version 2.1.234 or later [3]. While it waits, two lines at the bottom of the session read Usage limit reached · limit resets 3:45pm and Continuing automatically at 3:45pm · esc to cancel [3]. At the reset, Claude Code prompts Claude to continue the task from where it stopped, without resending the developer's last message [3].

Anthropic lists the conditions. If the computer slept for more than what Anthropic puts at 30 minutes and the limit reset while it slept, the developer presses Enter to continue [3]. The continued task runs like any other turn and still asks for permissions, so it can stop on a prompt while the developer is away [3]. If it reaches the limit again, Claude Code starts the wait again on its own at most twice in a row, then stops and shows Automatic continue stopped after repeated usage-limit hits · /rate-limit-options to try again [3]. Claude Code does not start the wait on its own in a Remote Control or agent team teammate session, when the reset is more than 24 hours away, which a weekly limit can be, or after an Opus or Sonnet limit while the developer runs a model outside that family; in those cases it opens a menu once per reset window, and the developer picks Wait here, then continue automatically [3]. It offers no wait at all in background sessions and -p runs, on API keys, cloud providers and usage-based billing, where usage is metered per request and there is no reset, or through an LLM gateway without a saved claude.ai login [3].

The organisation-wide control is the autoContinueAtUsageLimit setting. Anthropic's cost documentation says that to control for a fleet whether Claude Code starts the wait on its own, an administrator sets it in managed settings [1]. The settings reference, read on 4 October 2026, gives its scope as user or managed: Claude Code reads it from user settings, the --settings flag and managed settings only, and when none of those sets it, a project or local settings file that sets it turns the feature off [4]. Its default is true. It appears in /config as Continue automatically at usage limit, and Claude Code hides that row while managed settings or --settings set the key [4]. A developer can turn it off with the command form /config autoContinueAtUsageLimit=false, but Anthropic says that form cannot turn it back on, because the setting grants unattended execution [3]. The page on settings an administrator can enforce places this setting beside the other managed ones.

What to do when a developer asks

Reveneau's recommendation, from Anthropic's list, is to ask the developer to read the message aloud and to act on the first words. You've hit your session limit or weekly limit means waiting or usage credits, and the window restores itself at the reset time shown. You've hit your Opus limit means a model switch. You've hit your individual, org's or team's limit means an administrator raises a figure at Organization settings > Usage. A lowercase spend limit reached means the gateway operator. A context warning means the developer's own habits, and the sister guide on how to reduce Claude Code token usage is the page to send. Decide once, in managed settings, whether sessions should wait and continue on their own, so that every developer in the organisation gets the same behaviour.

Reveneau is an AI software development consultancy. All of its code is written by AI, and every change must pass an eval suite, a set of automated tests written from the specification, before release, so token use is a running cost of every Reveneau build. Reveneau is independent of Anthropic, and every message, version number and setting on this page comes from Anthropic's own documentation read on 4 October 2026. If you want help writing a limit response guide for your administrators, contact Reveneau.