Claude Code launches automatic mode! AI helps you determine execution permissions, so development processes no longer require individual approvals.

Anthropic has launched Claude Code Auto Mode, which uses a classifier to automatically assess tool call risks, bridging the gap between approval and skip permissions, and speeding up development.

On March 24, Anthropic announced the official release of Claude Code Auto Mode. This new feature allows developers to avoid manual approval every time a file is written or a Bash command is executed, and also eliminates the need to fully skip permissions—Claude automatically determines whether each operation is safe and whether to execute it.

How it works: Classifier evaluates risk before each tool call

The core of Auto Mode is a classifier that automatically reviews each tool call before execution. The assessment results fall into two categories:

• Safe operation: automatically executed without developer intervention
• High-risk operation: directly blocked, with Claude taking an alternative execution path

Anthropic emphasizes that Auto Mode reduces risk but cannot eliminate it entirely, and recommends using it within isolated environments.

How to enable and applicable plans

Auto Mode is currently available in “Research Preview” on the Team plan, with Enterprise and API access expected to be rolled out in the coming days.

Activation command:

claude --enable-auto-mode

After enabling, pressing Shift+Tab allows switching between modes.

Bridging the gap between “needs approval” and “completely skip”

Previously, Claude Code offered two extremes: manually approving each operation or using --dangerously-skip-permissions to skip entirely. The former frequently interrupted development flow, while the latter posed high risks. Auto Mode creates a middle ground—AI automatically evaluates, allowing safe operations to proceed automatically and blocking dangerous ones.

The community response has been enthusiastic, with many developers saying, “The approve/deny process has always been a bottleneck, not the model itself.” Some also point out that this is a key step in Claude Code moving from a “co-pilot” to an “operator”—trust issues replacing technical capability as the main obstacle.

Anthropic’s Impressive Release Pace in 2026

Auto Mode exemplifies Anthropic’s rapid iteration this year. Community timelines show that from January to March 2026, Anthropic has rolled out over twenty features including Claude Cowork, Opus 4.6, Sonnet 4.6, PowerPoint/Excel integrations, Claude Code Channels, Computer Use, Dispatch, and now Auto Mode—averaging a major update roughly every three to four days.

Further reading
Complete analysis of .claude/ folder

Figma MCP + Claude Code

Claude Desktop control features

Complete guide to Claude Code Skills

• This article is reprinted with permission from：《Chain News》
• Original title: “Claude Code Launches Auto Mode: AI Automatically Determines Execution Permissions, No More Manual Approvals in Development”
• Original author: Elponcrab