이미지: X — 프론티어랩
Summary
- Claude Code's desktop app has added an "auto-resume" checkbox that automatically continues work once a usage limit is reached
- With the checkbox enabled, work resumes from where it stopped the moment the limit resets, with no manual intervention needed
- The published screenshot showed a long-running migration task moving a payment client from v2 to v3
- 기능
- Claude Code 데스크톱 앱 '자동 재개(auto-continue)' 체크박스
- 발표
- ClaudeDevs 계정, 2026년 8월 13일
- 작동 방식
- 사용량 한도 도달 시 작업 중지, 한도 리셋되면 자동으로 이어서 재개
- 예시 작업
- 결제 클라이언트 v2→v3 마이그레이션, 61개 파일 214개 호출 지점
- 화면 표기
- 모델 Opus 5, 요금제 Max
- 관련 최근 업데이트
- 8월 9일 예산 상한·어드바이저 기능, 8월 12일 크롬 세션 연동
Work no longer stalls when limits are hit
Anyone who has handed a large refactor to an agent, stepped away, and come back to find it stalled for hours after hitting a usage limit will find this familiar. Anthropic's coding agent tool, Claude Code, has added a new "auto-resume" checkbox to its desktop app aimed at exactly this problem. When enabled, it automatically continues the interrupted work once the usage limit resets. The X account ClaudeDevs, which covers development news, announced the feature on August 13.
What the published screenshot shows
The accompanying screenshot showed a real task in progress. When the user instructed the agent to "migrate all call sites to the v3 payment client and remove the v2 shim, but keep tests passing," the agent read 38 files, searched for 6 patterns, and began processing 214 call sites across 61 files module by module. At the bottom of the screen, a progress indicator read "Migrating... 43m 10s," with the model listed as Opus 5 and the plan as Max. Work at this scale routinely takes hours, and until now, if a usage limit was hit partway through, a person had to come back and manually press resume.
How to turn it on
The original announcement didn't specify an exact menu path, but based on the published image, the feature appears as a checkbox within the task screen of the Claude Code desktop app. The confirmed process is as follows:
- Instruct the agent to perform a task as usual in the Claude Code desktop app.
- Work pauses once the usage limit is reached.
- With the auto-resume checkbox enabled, work resumes automatically from where it stopped as soon as the limit resets, with no further action needed.
If the checkbox is left unchecked, it appears the previous behavior still applies — the user must manually trigger resumption after the limit resets.
Who can use it
The announcement specifically referred to the Claude Code "desktop" app. Whether the same feature applies to the web or mobile versions was not confirmed in this announcement. The published screenshot showed the Max plan, but the original announcement did not separately address support for other plans.
Part of a recent string of updates for long-running tasks
This feature is in line with a series of updates Anthropic has rolled out over the past few weeks. On August 9, a hard budget cap was added to Claude Managed Agents, pausing them in a "budget_reached" state without issuing new requests once the budget is exhausted. On August 12, Claude in Chrome conversation sessions gained the ability to carry over across desktop, web, and mobile, starting with the Max and Team plans.
| Announcement date | Feature | Key details |
|---|---|---|
| August 9 | Budget cap and advisor | Automatically pauses when budget is exhausted; can request advice from a more capable model |
| August 12 | Chrome session carryover | Browser conversations carry over to desktop, web, and mobile |
| August 13 | Auto-resume | Automatically resumes interrupted work when the limit resets |
How long and how reliably an agent can work unsupervised is one of the central battlegrounds in the coding-agent race. Budget management, session carryover, and auto-resume each address the problem from a different angle, but they all point in the same direction — reducing the need for a human to stay glued to the screen.
What actually changes
Until now, handing off hours-long tasks like large migrations meant someone had to stay available to re-issue instructions whenever a usage limit was hit. The auto-resume checkbox eliminates that repetitive step. A refactor left running overnight that hits its limit will now simply pick back up the moment the limit resets — meaning users can check the results the next morning instead of manually restarting it. That said, the scope of support outside the desktop app and any plan-specific conditions are not yet fully clear from this announcement alone.
