# Auto Mode

> A second model checks each step. You are rarely asked.

**Field:** Artificial Intelligence  
**Category:** Claude Code  
**Also known as:** --permission-mode auto, auto mode on, Permission classifier

## Think of it like

It is like a guard at the door who checks each guest for you.

## Definition

Auto mode lets Claude Code run most actions without asking you. In your place, a second model checks each action before it runs. It blocks actions that go past what you asked for, or that reach places it does not know. It also blocks ones that look like a prompt injection. Claude Code still asks you about anything your own ask rules name. New sessions in the terminal start in this mode unless you set another.

## Example

Jack leaves a long refactor running in auto mode. The checker lets the edits and test runs through, but blocks a command that would push to the main branch.

## Example code

```bash
claude -p "fix the lint errors" --permission-mode auto
```

Run a script job with checks but no prompts.

## Why it matters

Long tasks keep moving without a wall of prompts, while risky actions are still caught.

## Common mistake

Treating it as a sandbox. The checker can be wrong, so keep deny rules and backups for things that must never happen.

## Three principles

- **A second model.** Checks each action.
- **Few prompts.** Safe ones just run.
- **Your ask rules.** Still ask you.

## Related terms

- [Permission Mode](https://learn.benibix.com/term/ai/permission-mode)
- [Prompt Injection](https://learn.benibix.com/term/ai/prompt-injection)
- [Permission Rule](https://learn.benibix.com/term/ai/permission-rule)
- [Claude Code Sandbox](https://learn.benibix.com/term/ai/claude-code-sandbox)

## Told apart from

- [Auto Mode vs Bypass Permissions](https://learn.benibix.com/compare/auto-mode-vs-bypass-permissions-mode): People often think auto mode skips the checks, like bypass mode. Both stop the prompts, but only one keeps a guard. In auto mode, a second model checks each action and blocks the risky ones. In bypass mode, nothing checks most actions at all. Auto mode is meant for daily work. Bypass mode is only for a container you can throw away.
- [Plan Mode vs Auto Mode](https://learn.benibix.com/compare/plan-mode-vs-auto-mode): People think plan mode and auto mode are two ways to let Claude get on with it. They answer different questions. Plan mode decides what Claude may do yet: it can read and think, but it makes no edits until you approve a plan. Auto mode decides who says yes: Claude acts, and a second model checks each step instead of asking you.

## Taught in

- [Claude Code Permission Modes](https://learn.benibix.com/learn/ai-claude-code/claude-code-permission-modes): Claude Code, 5/16
- [Setup and Config Commands](https://learn.benibix.com/learn/ai-claude-code-commands/setup-and-config-commands): Claude Code Commands, 9/17

![Each action goes to a second model that checks it. Safe ones run, risky ones are blocked or sent to you.](https://learn.benibix.com/diagrams/cc-auto-mode.svg)

---

Source: https://learn.benibix.com/term/ai/auto-mode
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-19
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
