Claude Code Now Gets a Grace Period at the Usage Limit

Anyone who has run Claude Code on a subscription knows the moment: the agent is halfway through a refactor, the five-hour usage limit arrives, and the session stops mid-edit, leaving a half-applied change and a build that no longer compiles. Anthropic has quietly softened that edge. A Claude Help Center article added in late September describes a “wrap-up allowance”: when the limit is reached while a response is already under way, Claude Code may keep working for a short while to reach a logical stopping point, and the status line reads “Usage limit reached · wrapping up”.

Who gets it

Four conditions all have to hold. The account must be on a Pro or Max plan, or hold a Premium seat on a Team plan, and be in good standing. Claude Code must be version 2.1.277 or later, in the terminal, an IDE or the desktop app. The account and its organisation must clear minimum age thresholds that Anthropic does not publish and says may change. And the account has to be part of a gradual rollout, so not everyone who qualifies on paper has it yet. Usage through an API key, through third-party cloud providers, and ordinary chats in the Claude apps are all excluded.

How it works

The allowance only covers a response that is already running when the five-hour limit hits; it never lets a new message start. Pro subscribers get it at most once per weekly usage period. On Max and Team Premium it can apply every time the five-hour limit is reached, but always inside the weekly cap. Whatever the wrap-up uses counts against the weekly limit rather than adding to it, and the response ends when the work is finished, when the allowance runs out or when the user interrupts. It applies automatically, and it cannot be saved up or carried over.

Anthropic does not say how big the allowance is, only that it is small, capped and varies by plan. For subscribers who have turned on usage credits, the included allowance is spent first and the response then carries on at standard credit rates, subject to whatever spending limits are already set.

The help article is firm about what the feature is not. It is not a credit or a promotion, it has no cash value, it is not part of a plan’s advertised limits, and a missing or reduced allowance is not grounds for a refund. Anthropic reserves the right to change the amount, frequency or eligibility, or to drop it entirely without notice, and can withhold it over fraud, abuse or attempts to get around the limits.

What it looks like

We ran into it ourselves on 6 October. A long Claude Code session at BitsMinds reached the five-hour limit in the middle of a task, and instead of cutting off, the status line switched to “Usage limit reached · wrapping up” and Claude carried on until it had finished the work in hand. There was nothing to claim or confirm; it simply kicked in, as the help article describes.

The change follows Anthropic’s 50% increase in Claude Code limits in May, but it targets a different complaint. Bigger limits mean the wall arrives later; the wrap-up allowance means that when it does arrive, the code on disk is more likely to be in a state worth coming back to.