Anthropic turns on Claude Code’s auto mode by default

Claude Code will soon edit files and run commands without asking for your approval at every single step.

Starting August 14, Anthropic is making auto mode the default setting for Pro, Max, and Team accounts, according to TechCrunch. First tested in March, the feature lets the coding agent execute multi-step tasks continuously, pausing only when an action is irreversible, destructive, or targeted outside your local environment.

Why it matters: Humans are bad at clicking approval popups. In an Anthropic study with 1,053 paid testers, users approved 97% of permission prompts out of habit, catching only 13.6% of harmful actions. Auto mode caught 89%.

To back up the push, Anthropic added safety layers like prompt injection screening and customizable hard deny rules to stop data exfiltration. Claude Code Head Boris Cherny posted on X that his team has used auto mode exclusively for months and can't imagine returning to permission prompts.

Turns out the best way to keep AI from breaking your project is taking away your sleepy clicker finger.