Anthropic is making auto mode the default in Claude Code for Pro, Max, and Team plans starting August 14. New sessions on these plans will automatically run in auto mode, though users who have set a different default may receive a one-time prompt to switch. Those with pinned defaults will see no changes. The auto mode classifier will no longer incur additional charges for Pro, Max, and Team users, effective immediately. Auto mode remains opt-in for Claude Enterprise, the Claude API, and various cloud platforms, with plans to expand the default across all platforms in the coming month.

Auto mode routes each tool call through a classifier designed to block irreversible, destructive, or environment-exiting actions. When blocked, Claude typically finds safer alternatives or asks for user approval. After three consecutive blocks or twenty total in a session, it falls back to manual approvals. Testing showed auto mode matched or outperformed manual review, with 1,053 paid testers approving only 13.6% of dangerous commands compared to auto mode's 89% block rate.

Research across internal red-teaming, third-party evaluations, and real production sessions demonstrated auto mode's superior safety. Among flagged sessions, manually approved workflows contained unintended harm more than twice as often as auto mode sessions. At production-level severity (7+), 6.3% of manual sessions had harmful unrequested actions versus 2.4% for auto mode. Collaboration with Apollo Research further strengthened the classifier through synthetic attack testing.

Teams at Adobe, Nuro, Gusto, and Garner Health already use auto mode as their production default. Auto mode users among Teams and Enterprise adopters ship approximately 25% more PRs, enabling longer uninterrupted autonomous work.