Auto mode is now the default in Claude Code

Claude Code will soon switch to auto mode by default for Pro, Max, and Team plans, starting August 14. This change allows for longer-running autonomous work and helps catch dangerous commands that manual review might miss. Auto mode classifies tool calls to block irreversible, destructive, or external actions. When blocked, Claude finds a safer alternative or asks for confirmation. If unable to proceed for three consecutive blocks or twenty across a session, it falls back to manual approvals.

Auto mode uses extra tokens per tool call, now free for Pro, Max, and Team users. It remains opt-in for Claude Enterprise, Claude API, Claude Platform on AWS, Amazon Bedrock, Google Cloud's Agent Platform, and Microsoft Foundry. Over the next month, Claude Code plans to make auto mode the default across all these platforms and stop charging for classifier overhead.

Auto mode balances user convenience and safety. It reduces the need for frequent permission approvals, which can become habitual and lead to overlooking important security decisions. In testing, auto mode matched or outperformed manual review in blocking dangerous commands. It also allows longer-running tasks, making Claude suitable for larger projects. Teams at Adobe, Nuro, Gusto, and Garner Health already use auto mode in production.

While Claude's safety data and customer results support the change, admins can still set a different default if desired.