The feature acts as a compromise between careful guidance and excessive autonomy.
Anthropic introduced an “auto mode” for Claude Code, a tool enabling AI to make permission-level decisions for users. This feature offers vibe coders a safer alternative between constant guidance or too much autonomy. Claude Code can act on users’ behalf, a useful but risky function, as it might perform unwanted actions such as deleting files, sending sensitive data, or executing harmful code. Auto mode aims to prevent this by flagging and blocking risky actions before execution and allowing users to intervene.
Currently, auto mode is available as a research preview for Team plan users, with plans to extend access to Enterprise and API users shortly. Anthropic notes the tool is experimental and advises using it in isolated environments.