Starting August 14, Anthropic will enable Auto Mode by default in Claude Code for subscribers on Pro, Max, and Team plans. In this mode, the assistant decides on its own whether an operation is safe to execute — relying on a built-in security classifier instead of requesting confirmation for every action. The change responds to notification fatigue: per company statistics, developers approve 97% of requests, having formed a habit of clicking “agree” without looking. Security evaluation consumes extra tokens, but Anthropic will make them free for higher-tier users; enterprise clients, API users, and cloud platform customers must still enable the feature manually for now, though Auto Mode is planned to become the standard for everyone.

Related: Anthropic Measures AI Agent Autonomy in Real-World Use, Anthropic Launches Claude Code Security

Claude Blog: Auto Mode Default in Claude Code