How to Use Claude Auto Mode Default in Apps

Claude sessions on the web, desktop, and mobile apps now start in auto mode by default for Pro, Max, and Team plans. Anthropic announced the change on October 22, 2026. Starting the week of October 26, an automated classifier reviews each action instead of prompting you for approval every step.

This matches the default Anthropic already set in Claude Code on August 14, 2026. Safe actions proceed. Risky ones get blocked. You can switch back to manual approval at any time from the session controls.

What Auto Mode Does

Auto mode sits between full manual approval and skipping all permission checks. A separate classifier model examines each proposed tool call or action before it runs. It looks for signs of data exfiltration, mass file deletion, malicious code, or other destructive behavior.

In Claude Code testing, the classifier caught 89% of dangerous commands. Manual review caught only 14%. The classifier does not see Claude’s internal reasoning. It evaluates the action against your messages and context.

If the classifier flags something, Claude is redirected to try a different approach. Repeated blocks can still surface a permission prompt. Auto mode does not remove all risk. Anthropic recommends using it in isolated or sandboxed environments when possible.

How to Switch Modes in Claude Apps

On the web, desktop, and mobile apps, new sessions start in auto mode for eligible plans. To return to manual approval:

1. Open the active session.
2. Tap or click the permission or settings control in the message input area.
3. Select Manual (or the equivalent label for step-by-step approval).

The change applies to the current conversation. You can toggle it back to auto later. Enterprise plans remain on manual by default for now.

For Claude Science sessions, auto mode requires the sandbox to be enabled and organizational allowance on Team or Enterprise plans.

Using Auto Mode in Claude Code

Claude Code already defaults to auto mode for new Pro, Max, and Team sessions. To cycle modes during a session, press Shift+Tab. The sequence moves through default (manual), acceptEdits, plan, and auto.

To pin auto mode as your starting preference, add this to your user settings file at ~/.claude/settings.json:

{
  "defaultMode": "auto"
}

Organization-managed settings can disable auto mode entirely with "disableAutoMode": "disable". Project-level settings cannot force auto mode on users.

Practical Tips for Daily Use

Start with low-stakes tasks while you learn what the classifier allows. Keep destructive operations such as bulk deletes or external sends in manual mode until you trust the boundary.

Combine auto mode with clear goals in your prompt. For coding work, pair it with the approaches in our guide on using Claude Opus 5.5 for agentic coding.

Monitor token use. Auto mode can add a small overhead from the classifier checks. On Team and Enterprise plans, admins can review usage and set organization-wide defaults.

When to Prefer Manual Approval

Use manual mode when the task involves sensitive data, production systems, or irreversible actions. The classifier is strong but not perfect. Ambiguous intent or incomplete context can lead to an allowed action you did not expect.

You can switch mid-session without losing progress. The mode applies to the rest of the conversation unless you change it again.

Conclusion

Claude auto mode reduces constant approval prompts while keeping a safety classifier in the path. It is now the default across web, desktop, and mobile apps for Pro, Max, and Team users, matching the experience already live in Claude Code. Switch back to manual any time from the session controls if you prefer tighter oversight.

Test it on a non-critical task first. Then decide which mode fits each kind of work.

Sources

• Anthropic Claude X post, October 22, 2026
• Anthropic auto mode documentation and Claude Code blog
• Claude Code permission modes documentation