Anthropic launched auto mode for Claude Code on March 24, 2026, currently available for Team users.
Default Claude Code asks permission for every action, slowing down long tasks and breaking development flow.
Using --dangerously-skip-permissions removes safety checks completely, which can lead to serious mistakes or unintended system damage.
Auto mode allows Claude to work independently while still keeping safety checks active for controlled automation.
A built-in classifier reviews each tool action before execution, ensuring only safe operations run automatically.
Dangerous actions like deleting files, leaking data, or harmful code execution are automatically blocked by system.
Claude pauses only when an action is risky and cannot be safely completed without your manual approval.
Choose between default, auto mode, or skip-permissions depending on your need for safety, speed, and control.
Run claude --enable-auto-mode in terminal, then press Shift+Tab anytime to switch modes during session.
Use auto mode in containers, virtual machines, or development environments instead of production systems for safety.
Auto mode is a research preview, and ongoing improvements depend on real-world feedback and usage patterns.