Claude Code Auto Mode

Auto mode is a permission mode in Claude Code, Anthropic's agentic coding tool. Instead of asking the user to approve each action, every tool call goes through an automated classifier. It is intended to block actions that are irreversible, destructive, or aimed outside the user's environment. Safe actions proceed, and when something is blocked Claude usually tries a safer route or asks the user directly. After three blocks in a row, or twenty in a session, Claude Code falls back to manual approvals.

Anthropic says auto mode became the default for new Claude Code sessions on Pro, Max and Team plans starting August 14. An October 22, 2026 update says it is rolling out as the default in most Claude apps on those plans. Users who set a different default may get a one-time prompt, and users with a pinned default see no change. Permission rules still apply before the classifier, but broad allow rules that grant arbitrary code execution (for example python:*) are set aside while in auto mode. Settings files are not modified.

Anthropic reports that it tested auto mode with internal and third-party red-teaming, prompt-injection evaluations, a study of 1,053 paid testers, and analysis of production sessions. It says auto mode matched or outperformed manual review on every measure it tested. These are vendor-reported results. Anthropic also says the classifier's extra token overhead is no longer charged on Pro, Max, Ultra and Team plans. On Enterprise, the API, Amazon Bedrock, Google Cloud and Microsoft Foundry it remains opt-in for now, and Enterprise admins can make it the default through managed settings.