Anthropic launches auto mode for Claude Code

Anthropic has introduced “auto mode” for its Claude Code tool, a feature designed to reduce the need for developers to manually oversee every action taken by the AI. Currently in research preview—meaning the feature is available for testing but is not yet a finished product—the update aims to eliminate the need for developers to babysit every action. According to Anthropic, the tool is designed to let the AI decide which actions are safe to take on its own, balancing speed with control. This reflects a broader shift across the industry, as AI tools are increasingly designed to act without waiting for human approval.

To manage this autonomy, the feature uses built-in AI safeguards to review each action before it runs. These safeguards check for “risky behavior” that the user did not request, as well as signs of “prompt injection.” Prompt injection is a type of attack where “malicious instructions” are hidden in content that the AI is processing, which can cause the system to take unintended actions. By filtering these actions, the safeguards aim to prevent “risky and unpredictable” outcomes. Safe actions are allowed to proceed automatically, while those flagged as risky are blocked.

According to Anthropic, the auto mode feature currently only works with its Claude Sonnet 4.6 and Opus 4.6 models. Because of the potential risks of letting the model run unchecked, the company recommends that developers use the new feature in “isolated environments.” These are sandboxed setups kept separate from production systems to limit potential damage if something goes wrong.

The launch of auto mode follows similar moves by industry peers like GitHub and OpenAI, which have also been developing autonomous coding tools that can execute tasks on a developer’s behalf. Anthropic plans to roll out the feature to Enterprise and API users in the coming days.

Why it matters

Anthropic’s new “auto mode” aims to eliminate the need for developers to babysit every AI action by allowing the model to decide which actions are safe to take, balancing speed with control.