Anthropic’s Claude Code Gets ‘Safer’ Auto Mode

Anthropic has added auto mode to Claude Code. This mode lets the model carry out actions for you, aiming for a balance between giving you complete control and letting the system act on its own.

• Before any risky action—such as deleting files, sending data out, or running hidden code—the system checks for danger.

• If a risk is found, the action is stopped. The system then either tries again safely or asks you to step in.

The feature is currently a research preview available only to Team‑plan customers. Anthropic plans to open it up to Enterprise and API users later.

The tool is experimental, does not eliminate all risk, and should be used only in isolated environments.