Ever clicked "Allow" so many times you stopped reading what you were allowing?
Anthropic noticed too.
Starting this week, Claude on the web, desktop and mobile runs in auto mode by default on Pro, Max and Team plans.
Here's how Anthropic put it: "Rather than asking you to approve each step, an automated check blocks risky actions."
So Claude just keeps working, and a safety check watches each action as it happens.
(Auto has been the default for new Claude Code sessions on Pro, Max and Team since August 14. Now the chat apps catch up.)
Right after this: the three modes and what each one does, the connector setting that matters more than the mode, and when to switch back to manual.
Now a quick word from today's sponsor.
285 Free Prompts to Save Time
Your inbox isn't the problem. Neither are your meetings. It's the hours you spend on work AI could do in minutes.
285 copy-paste prompts: built for email, meetings, admin, and delegation. Each one has a clear role, a specific task, and a defined output.
No setup. No prompt engineering. Just copy, paste, and get your time back. Free when you subscribe to Mindstream.
Right. Back to it.
The three modes
You'll find them in the mode selector in the box where you type your message. You can switch at any time.
Manual ("Manually approve"). Claude pauses and asks before each action. You click Allow or Deny. This used to be called "Ask before acting."
Auto ("Automatically approve"). Claude keeps going without asking about every step. A safety check reviews each action and blocks the unsafe ones. If Claude keeps hitting blocks, it goes back to asking you.
Skip ("Skip all approvals"). Claude doesn't pause, and nothing checks its actions. This used to be called "Act without asking."
Auto is the new default on Pro, Max and Team. Enterprise stays on Manual by default.
On Team and Enterprise plans, an admin can switch Auto off. If they do, it won't show up in your mode selector.
The connector setting that matters more than the mode
Every connector (an app you've hooked up to Claude) has its own permission: "Always allow," "Needs approval" or "Blocked."
In Auto mode, those labels work differently:
Read-only tools are approved automatically.
For tools that write or delete, and for anything set to "Needs approval," Claude decides for itself.
Only "Blocked" is always refused.
So if there's a tool you never want Claude using on its own, "Blocked" is the only setting that guarantees it in Auto mode.
When to switch back to manual
Anthropic's own advice: for work with real consequences, "stay close and review what Claude does," or switch back to "Manually approve."
A simple rule:
Auto for research, drafts, summaries and sorting files in a test folder.
Manual for anything that sends, deletes, pays or touches a client's files.
Skip for nothing that matters. Nothing checks its work.
Do this today
Open Claude and check which mode your next chat starts in.
Go through your connectors and find anything that can send or delete (email is the big one). Either set it to "Blocked," so Claude can't use it at all, or leave it on "Needs approval" and do that work in Manual mode, where Claude has to ask you first.
Try one low-stakes task in Auto, like "summarize the files in this folder and suggest a better naming system," and watch what it does without asking.
Make it stick
Decide once which mode fits most of what you use Claude for, and start there.
Then switch per task, not per click.
Letting AI work without asking every two minutes only pays off when it already knows your rules: what it may touch, what it may send, what needs you. That's the whole idea behind The One-Person Company Playbook.
It's the complete setup for running a business where AI does the work of a full team. Inside:
A company folder with your brief, pipeline tracker and invoice log already laid out, so any AI picks up where you left off
Thirty AI roles across six departments, from sales to finance, ready to use
Six jobs that run on a clock: the morning briefing, follow-ups, invoice checks and more
The step-by-step path from picking a business model to your first paying customer
Works on Claude, ChatGPT, Gemini or Grok, whichever you already pay for.
The catch: it takes about eight hours over a weekend to set up. After that, it's yours. $97, one-time (down from $297).
Talk tomorrow,
Zephyr
P.S. Are you keeping Auto on, or going back to Manual? Hit reply and tell me which, and why.