Claude Code gets a wrap-up budget at the 5-hour limit

If Claude Code runs out of its 5-hour budget in the middle of an edit, it will now try to find a graceful stopping point instead of cutting off there, according to a post from the ClaudeDevs account on X. To finish what it can, it gets a small, fixed allowance, and that allowance comes out of your weekly limit.

At a glance

• How often you get the wrap-up depends on your plan during the rollout: Pro subscribers get it once a week, while Max and Team Premium get it every time a 5-hour session limit is hit.
• The extra work has a cost: it draws on a small, fixed allowance taken from your weekly limit, and if you need to keep going past the wrap-up, you can continue with extra usage.
• Anthropic has not said how large the allowance is, only that it is small and fixed, and Claude Code promises only to try to find a stopping point, not to finish the task.

If you missed the earlier rounds: according to Truefoundry, usage limits came in because a small share of power users were running 24-hour sessions, sharing credentials across teams and burning thousands of dollars of compute on $20 subscriptions. The same outlet reports that Anthropic reduced 5-hour limits during weekday peak hours (5–11 AM PT) starting March 2026, and acknowledged on March 31 that users were hitting limits faster than expected.

Pro gets the wrap-up once a week, Max and Team Premium at every 5-hour limit

How often you get the wrap-up during the rollout depends on your plan. On Pro, it is available once a week. On Max and Team Premium, it kicks in each time a 5-hour session limit is hit. On every plan, whatever Claude Code does while wrapping up comes from a small, fixed allowance pulled from your weekly limit, so the extra work is still paid for.

If the wrap-up is not enough, the post points to extra usage. According to Claude's help pages, usage credits switch you to consumption-based pricing at standard API rates once you reach your plan's limit, and they cover both Claude conversations and Claude Code terminal usage. The same pages say credits do not change the five-hour reset, and that in certain jurisdictions such as Japan, credits expire six months after purchase starting September 10, 2026.

The wrap-up is charged to the weekly limit, not the 5-hour one

Claude Code runs on two clocks. According to Truefoundry, there is a 5-hour rolling window for short-term activity and a weekly cap on active compute hours, and Claude Code, Claude.ai chat and Cowork all draw on the same bucket. SessionWatcher notes that Anthropic's docs let the weekly limit for Opus reset on its own schedule, separately from all other models.

The weekly layer has moved recently too. Explainx reports that the ClaudeDevs account announced the temporary 50% weekly limit boost would end September 14, 2026, and be replaced the same day by a permanent 25% raise over each plan's baseline. The outlet works that out as a net cut of about 17% from the promo-inflated capacity.

Think of a phone plan with both a daily cap and a monthly one. The wrap-up lets you run slightly past today's cap by drawing a small amount from the month. You get to finish the call, and the monthly balance ends up a little lower.

Claude Code gets only a small, fixed allowance to finish what it can

The wrap-up budget has to cover whatever Claude Code does in its last few steps, and those steps are not cheap. According to Sitepoint, Claude Code does not send one prompt and wait. Each interaction is a multi-turn conversation that carries the system prompt, the accumulated history, file contents and tool-use tokens, so each request uses more tokens as the session goes on.

Sitepoint puts a seemingly simple "edit this file" command at between 50,000 and 150,000 tokens in a single API call once the full context window is assembled. By the same outlet's estimate, a 10-file project might use 50k tokens per request, while a 500-file monorepo can exceed 500k.

SessionWatcher lists what makes either limit drain faster: message length, file attachment size, current conversation length, tool usage such as research or web search, model choice, effort level, and artifact creation and usage.

Anthropic has disclosed nothing about the allowance beyond "small and fixed", and Claude Code only promises to try. Judging by Sitepoint's per-edit figures, a fixed allowance will likely go further in a small project than in a monorepo, and the post does not say what a Pro user gets at a second limit in the same week. In our view, charging the wrap-up to the weekly limit is a sensible design, because the cost shows up in a number you already watch.

What the rollout tiers leave open. The post describes the Pro and Max/Team Premium split as how things work "as we roll this out", and no date has been given for when the rollout ends or whether the tiers will change. The allowance size is also unpublished, so your own next limit will be the practical test: whether Claude Code lands on a clean stopping point, and how much of your weekly limit it spends getting there.