> ## Documentation Index
> Fetch the complete documentation index at: https://docs.trysixth.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Auto-Approve

> Choose which actions Sixth can take without asking.

Auto-approve lets Sixth take certain actions without stopping for your OK. Anything not auto-approved shows an [approval card](/features/chat-interface#approval-cards) in the chat, which you answer with **Enter** (approve) or **Esc** (reject).

## Opening the settings

Click the **shield** in the bottom row of the message box. Next to it is a short summary of what's currently auto-approved, for example **Read, Safe Commands**, **All**, **None** or **Off**. Click it to open the menu.

## Actions

| Action | What it allows |
| - | - |
| **Read project files** | Read files in your workspace. Its sub-option **Read all files** extends this to any file on your computer. |
| **Edit project files** | Modify files in your workspace. Its sub-option **Edit all files** extends this to any file on your computer. |
| **Execute safe commands** | Run terminal commands the model judges non-destructive. Its sub-option **Execute all commands** allows any command; use at your own risk. |
| **Use the browser** | Launch and interact with websites in a browser, and fetch web pages. |
| **Search the web** | Search the web for information without opening a browser. |
| **Use MCP servers & Workflows** | Use your MCP servers and connected [workflow integrations](/features/workflow-integrations) (Gmail, Drive, …), which may change files or call external APIs. |
| **Generate images** | Create AI images without asking. |

**Enable auto-approve** turns the whole feature on or off without losing your choices, and **Toggle all** switches every action at once. Star an action to keep it in your favourites for quick access.

New installs start with **Read project files** and **Execute safe commands** on.

Some low-risk actions never need approval, such as loading a skill, checking on or stopping a background agent, and waiting.

## Max requests

By default there's no limit: Sixth keeps working for as many requests as a task needs. If you'd like it to pause and check in, enter a number under **Max Requests**. Sixth stops after that many consecutive auto-approved requests and asks before continuing. Clear the field to go back to unlimited.

## Suggested setups

* **Careful:** read files only. You review every edit and command.
* **Balanced:** read files, edit project files and run safe commands. [Checkpoints](/features/checkpoints) let you undo anything you don't like.
* **Hands-off:** everything on, for prototyping or tasks you trust Sixth to run alone.

You can change these at any time, even mid-task.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.