ProductMarch 24, 2026

Claude Code Auto Mode Enters Research Preview

Anthropic is testing a safer auto mode for Claude Code that lets the model decide which actions can run without manual approval, a notable step toward more autonomous Claude workflows.

A more autonomous Claude Code

On March 24, 2026, TechCrunch reported that Anthropic began rolling out auto mode in Claude Code as a research preview. The feature is designed to let Claude decide which lower-risk actions it can perform automatically, instead of asking for approval at every step.

Why this is important

Auto mode matters because the biggest complaint about agentic coding tools is the tradeoff between speed and control:

  • too many confirmations make workflows slow
  • too few guardrails make them risky

Anthropic's answer is to insert an AI safety layer that evaluates whether an action looks aligned with the user's request and whether the context shows signs of prompt injection or other risky behavior.

Why it matters for Claude 5 speculation

If Anthropic ever decides to market a future system as Claude 5, many users will expect it to feel more like an operator than a chatbot. Auto mode is a concrete product signal in that direction.

This does not confirm a Claude 5 launch. What it does confirm is that Anthropic is shipping infrastructure for:

  • longer multi-step sessions
  • semi-autonomous execution
  • safer delegation inside coding environments

The real takeaway

The news here is less about branding and more about architecture. Anthropic appears to be building the behavior layer that would support a more agentic Claude generation, whether the next model is called Claude 5 or something else.

Sources

Ready to Experience Claude 5?

Try Now