Claude Code Auto Mode Becomes the Default on August 14 — What Changes and What to Configure

Starting August 14, new Claude Code sessions on Pro, Max, and Team plans will skip per-action approval prompts — a safety classifier handles them automatically, catching 89% of dangerous commands vs. 13.6% for manual review.

Auto mode has been opt-in since March. On August 14 it becomes the default for new sessions on Pro, Max, and Team plans. Anthropic's own testing explains why: users approve 97% of permission prompts reflexively — approval fatigue means manual oversight only catches 13.6% of dangerous commands. Auto mode's background classifier caught 89% of the same commands in the same study. Effective today, the classifier overhead is also free — no additional token charges.

What changes

Auto mode replaces per-action prompts with a classifier that evaluates each tool call in real time. If it flags something irreversible or out-of-bounds, Claude tries a safer path or asks for permission. After repeated blocks, the session falls back to manual approval.

Pinned defaults and managed admin settings are preserved. Claude Enterprise, API, Bedrock, Vertex AI, and Foundry remain opt-in for now — Anthropic plans a broader rollout there separately.

What to check before August 14

1. If you rely on per-action prompts for a specific workflow, pin your preference now:

// ~/.claude/settings.json
{
  "disableAutoMode": true
}

2. If your CLAUDE.md permission rules have drifted, run:

claude auto-mode reset

This restores the default auto-mode classifier behavior without touching your other settings.

3. Your --allowlist-file permissions are fully respected — anything you've explicitly permitted stays permitted; auto mode only handles the undecided cases.

After August 14, the Claude Code session header shows the active mode so you always know which approval flow is running.

Sources: Auto mode is now the default in Claude Code for Pro, Max, and Team plans — Anthropic · TechCrunch · Simon Willison's notes