Anthropic Launches Auto Mode for Claude Code

Anthropic has launched a new feature called auto mode for its Claude Code AI coding assistant, enabling the tool to independently approve routine actions while blocking potentially risky ones through built-in safeguards. The update, released as a research preview, aims to streamline developers' workflows by reducing the need for constant manual approvals during collaborative coding sessions.

Claude Code, Anthropic's AI-powered development tool, previously required users to approve each file write, shell command, or other operation, a process that often interrupted longer tasks. Developers could bypass this by using a command to skip permissions entirely, but that exposed systems to hazards like unintended file deletions or data exposure. Auto mode addresses this by employing an AI classifier to evaluate each proposed action before execution. Safe operations, such as routine code edits within a trusted directory, proceed automatically. Risky ones, including mass file deletions, sensitive data access, or suspicious commands, are blocked, prompting the AI to explore alternative approaches. If repeated blocks occur, the system escalates to seek user approval.