How do I turn away mode on and off?

You can press the chair icon in the chat header, or tell Claude 'turn on away mode,' 'I'm heading out,' or 'good night' to enable it. Say 'I'm back' or 'good morning' to turn it off; using the switch also triggers a report once Claude is free.

Does away mode ever approve actions on my behalf?

No. It only declines permission prompts after the wait time expires and never allows them, so the CLI's protections stay at whatever permission mode you are using.

Can I put a child session into away mode?

No. Only sessions a person is driving — the parent session and ordinary sessions not in a parent-child pair — can use it, because child sessions rely on the parent Claude to answer their questions and prompts.

Large Language ModelsAI Coding AssistantsZenn AI/ML29m agoOct 25, 2026, 16:00 JST

tanacode 2.4.0 adds away mode so Claude Code won't stall

Send on LINE

3 Key Points

What happened
tanacode 2.4.0's away mode lets Claude decide on its own and log choices while leaving irreversible or external actions on a 'human to-do' list; unanswered permission prompts are declined after a default 5 minutes.
Why it matters
Prompts come from the Claude Code CLI, not the model, so asking Claude to 'proceed without asking' never stopped the stalling — away mode declines prompts rather than allowing them, so safety settings stay as they are.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Get these free every morning

Context & Analysis

The stalling problem tanacode 2.4.0 targets comes from where permission prompts originate. Questions are a tool the model chooses to use, so telling Claude to decide without asking can reduce them. Permission prompts, by contrast, are issued by the local Claude Code CLI against its permission settings, which is why the author's own instruction to proceed without asking did not prevent overnight stalls.

The existing permission modes each fall short for an absent user. The auto mode stops and asks a person after three consecutive or 20 total blocked operations, and the bypassPermissions mode lets irreversible or external actions through without confirmation, with the official documentation recommending isolated containers or VMs. The dontAsk mode declines immediately, leaving no time to answer from a phone and no report of what was declined. Remote Control lets a user answer from a smartphone, but only while they are actually looking at it.

Away mode instead combines rules given to the model with time-limited answers from the app. Decisions and deferred items are recorded on checklists — 'items to confirm' and 'things only the user can do' — that Claude reviews on return, and the rules are re-sent with each user turn so they survive long conversations and context compaction. The author notes some limits: permission prompts cannot be declined while in terminal mode, and MCP servers that request human input are not covered.

FAQ

How do I turn away mode on and off?: You can press the chair icon in the chat header, or tell Claude 'turn on away mode,' 'I'm heading out,' or 'good night' to enable it. Say 'I'm back' or 'good morning' to turn it off; using the switch also triggers a report once Claude is free.
Does away mode ever approve actions on my behalf?: No. It only declines permission prompts after the wait time expires and never allows them, so the CLI's protections stay at whatever permission mode you are using.
Can I put a child session into away mode?: No. Only sessions a person is driving — the parent session and ordinary sessions not in a parent-child pair — can use it, because child sessions rely on the parent Claude to answer their questions and prompts.

Zenn AI/MLRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.