Anthropic has changed one of the most criticized parts of the Claude Code experience: what happens when a user hits the 5-hour usage cap.

According to the company’s developer account, @ClaudeDevs, Claude Code will no longer stop immediately at the limit. Instead, it will try to reach a clean stopping point, finish the step already in progress, and then halt. The feature is called Wrap-Up Allowance.
The update targets a problem many developers had described as especially disruptive. A task could be interrupted in the middle of a code change, after a function signature had been modified but before all call sites were updated, or while tests were still running. That often left users with a repository in a partial state and forced them to manually inspect what had already been written.
Developer @tochukwudev said this had been his biggest frustration with Claude Code, because each cutoff left behind incomplete changes that had to be sorted out by hand. Another developer, @PaineofGlass, said he used to work around the issue by asking Claude to save work in stages as the limit approached, or by raising the extra spending cap.
How the wrap-up allowance works
Anthropic says that when a user reaches the 5-hour cap, the system allocates a small fixed amount from the weekly quota so Claude can finish the current step before stopping. The interface shows the message: 「Usage limit reached · wrapping up.」

The company’s help documentation also sets clear limits on the feature:
- It only applies to the run already in progress. If the user sends a new message after hitting the cap, that request is still blocked until usage resets.
- It does not add extra total capacity. Any usage consumed during wrap-up is still deducted from the weekly limit.
- It does not guarantee the entire job will be completed. Once the wrap-up budget is exhausted and the current step ends, Claude will not begin the next step.
- Users can interrupt the process manually at any time.
Anthropic also says the allowance cannot be requested, accumulated, or carried over. It activates automatically when the account qualifies.
If the wrap-up budget is not enough, users can turn on extra usage and continue. The system first consumes the wrap-up allowance, then charges usage at the standard rate, still subject to the spending cap.
Access depends on the subscription tier
The rules differ by plan.

Pro users can use the feature at most once per week. Even if that one use does not consume the full wrap-up budget, it still counts as the week’s single use.
Max and Team Premium users can access it each time they hit the 5-hour limit, as long as their weekly quota has not been exhausted.
There are other requirements as well. Claude Code must be updated to version 2.1.277 or later. The feature works in the terminal, the IDE plugin, and the desktop Code page, but users must be signed in with a Claude subscription account. The account and the organization it belongs to must also have existed for a certain period, though Anthropic has not disclosed that duration and says it may change later.
The rollout is gradual, so not every account has it yet.

Anthropic says the feature is not available when Claude Code is used through an API key, Amazon Bedrock, or Google Cloud Vertex AI. It also does not apply to the standard Claude chat window.
As for the size of the wrap-up budget, the company only says it varies by plan and has a cap. No specific figures were disclosed.
The change comes after a weekly quota adjustment
Ten days earlier, on Sept. 14, Claude Code’s temporary 50% weekly quota promotion expired and was replaced with a permanent 25% increase.
Using a base quota of 100, that means the allowance moved from 150 back to 125, about 17% below the previous peak.
Anthropic initially said on X only that it had made a “permanent 25%” increase. After users criticized the wording and the post received a community note, the company deleted it and reposted with a clearer acknowledgment that, compared with the then-current level, the change represented a 17% reduction.

That context matters because Wrap-Up Allowance changes how Claude Code stops, not how much total usage a customer gets. The wrap-up budget still comes out of the same weekly pool.
Developers on X split over the update
Reaction on X was active and sharply divided.
@DataDiscovered said being cut off in the middle of editing was one of the worst failure modes in Claude Code, and called the wrap-up allowance one of Anthropic’s best fixes so far. @xDqtreats said he was moving from Codex to Claude Code in part because he liked this feature.
@nrajesh4u said he had reported the need for this long ago and believed thousands of others had raised the same issue. In that view, the fact that the feature arrived only now is itself part of the problem.

Others argued the change does not go far enough. @shamilkch said wrap-up only becomes truly meaningful if the next session can resume from a written checkpoint. Otherwise, he said, the fixed allowance mainly makes the stop look cleaner while remaining a timeout in substance.
Some comments focused on the 5-hour cap itself. @ysdevastro asked whether Anthropic could remove the 5-hour limit for Max users, saying that restriction was the main reason he had not switched from Codex.
The discussion shows that Codex and Claude Code are competing for the same group of developers, and that quota rules have become a real comparison point in product choice.
Anthropic changed the stopping point, not the total allowance
In practical terms, the update is narrow. Anthropic did not extend the 5-hour window. It did not raise the weekly quota. It did not make wrap-up usage free.

Still, the response was broadly positive because Claude Code is handling a different kind of workload than a standard chatbot. In a simple question-and-answer exchange, an interruption can often be fixed by asking again. In an agent-style coding workflow, Claude may read several files, edit code in multiple places, run tests, and iterate for tens of minutes in a single task.
In that setting, the state of the work is not contained in one reply. It sits in the middle of a chain of actions. Where the system stops determines whether the developer can continue from a usable point or has to spend time cleaning up the aftermath first.
Wrap-Up Allowance addresses that exact moment. It gives Claude a chance to finish the current step before the session ends.
The source material cited the @ClaudeDevs post on X and Anthropic’s help center documentation for Claude Code Wrap-Up Allowance. The original article also said it came from the WeChat public account Xinzhiyuan, written by ASI Qishilu.

