sakutto
Generative AI

Claude Code's Auto Mode Becomes the Default: What Runs and What Stops

Claude CodeAnthropicDevelopment environment
Claude Code's Auto Mode Becomes the Default: What Runs and What Stops

What auto mode is

Auto mode in brief

Reviewer
Not a person, but a classifier model (a separate AI)
Default
The starting mode on Pro, Max, and Team
Exception
Explicit ask rules always prompt

Auto mode is one of Claude Code's permission modes. For the basics, see the Claude guide.

A classifier model reviews in place of a person

In auto mode, actions run without a permission prompt each time. Instead, a separate AI called a classifier model reviews them beforehand and blocks anything that escalates beyond your request, targets unrecognized infrastructure, or looks driven by hostile content Claude read.

View official source →
"Auto mode lets Claude execute without routine permission prompts. A separate classifier model reviews actions before they run, blocking anything that escalates beyond your request, targets unrecognized infrastructure, or appears driven by hostile content Claude read. Explicit ask rules still force a prompt."(Eliminate permission prompts with auto mode section) — from the Claude Code official documentation

The starting mode on Pro, Max, and Team

On Pro, Max, and Team plans, auto mode is the built-in starting mode. That applies to terminal and VS Code extension sessions; claude -p, SDK usage, and the first session right after install start in Manual. It also requires v2.1.228 or later (v2.1.233 on Windows).

View official source →
"On Pro, Max, and Team plans, auto mode is the built-in starting mode."(Eliminate permission prompts with auto mode section)/"The built-in `auto` default requires Claude Code v2.1.228 or later on macOS, Linux, and WSL, and v2.1.233 or later on native Windows. On earlier versions, the built-in default is Manual."(Which mode a session starts in section) — from the Claude Code official documentation

What actually stops

Blocked by default (selected)

External execution
Fetching and running code, such as curl | bash
Production
Production deploys and migrations
Destruction
Mass storage deletion and force push
Permissions
Granting IAM or repository permissions

The documentation enumerates the line.

Representative blocks

The classifier trusts your working directory and the remotes configured for it when the session started, treating everything else as external. Blocked by default: downloading and executing code, sending sensitive data to external endpoints, production deploys and migrations (applying database structure changes), mass deletion on cloud storage, granting IAM (cloud permission management) or repository permissions, modifying shared infrastructure, irreversibly destroying pre-existing files, force push (overwriting history), git reset --hard, and terraform destroy.

View official source →
"The classifier trusts your working directory and the remotes that were configured for it when the session started."(What the classifier blocks by default section)/"Downloading and executing code, like `curl | bash`"/"Sending sensitive data to external endpoints"/"Production deploys and migrations"/"Mass deletion on cloud storage"/"Granting IAM or repo permissions"/"Modifying shared infrastructure"/"Irreversibly destroying files that existed before the session"/"Force push"/"`git reset --hard`, `git checkout -- .`, `git restore .`, `git clean -fd`, `git stash drop`, or `git stash clear`, which the classifier presumes would discard uncommitted changes"/"`terraform destroy`, `pulumi destroy`, `cdk destroy`, or `terragrunt destroy`, and applying a plan that destroys resources"(all from the Blocked by default list in the same section) — from the Claude Code official documentation

It does not make things safe

The documentation states plainly that auto mode reduces permission prompts but does not guarantee safety. It is positioned for work whose general direction you trust, not as a substitute for reviewing sensitive operations.

View official source →
"Auto mode reduces permission prompts but does not guarantee safety. Use it for tasks where you trust the general direction, not as a replacement for review on sensitive operations."(warning box in the Eliminate permission prompts with auto mode section) — from the Claude Code official documentation

Switching and disabling it

How to change it

During a session
Shift+Tab cycles in the CLI; VS Code and others use the mode selector
Change the default
permissions.defaultMode in ~/.claude/settings.json
Block it org-wide
Set permissions.disableAutoMode to disable in managed settings

None of this is fixed in place.

Switching in your own environment

The CLI cycles modes with Shift+Tab, and VS Code, Desktop, and claude.ai use the mode selector. To change the starting mode, use permissions.defaultMode — but since v2.1.142, auto does not take effect from a repository's .claude/settings.json or .claude/settings.local.json, so put it in ~/.claude/settings.json.

View official source →
"Cycle modes with Shift+Tab in the CLI or use the mode selector in VS Code, Desktop, and claude.ai."(page introduction)/"In Claude Code v2.1.142 and later, `auto` doesn't take effect from those files. Move it to `~/.claude/settings.json`."(Eliminate permission prompts with auto mode section) — from the Claude Code official documentation

Turning it off for an organization

To stop developers in an organization from using auto mode, set permissions.disableAutoMode to disable in managed settings. That removes auto from the Shift+Tab cycle, and a session started with --permission-mode auto starts in Manual instead.

View official source →
"Organization: on Team and Enterprise, auto mode is available by default. Administrators can turn it off for the organization by setting `permissions.disableAutoMode` to `"disable"` in managed settings."(Requirements block)/"To prevent developers from using auto mode, set `disableAutoMode` to `"disable"` in managed settings. This removes `auto` from the `Shift+Tab` cycle, and a session started with `--permission-mode auto` starts in Manual instead."(Auto mode on Bedrock, Agent Platform, or Foundry section) — from the Claude Code official documentation

When you want to read the documentation on your own machine as markdown with its headings and tables intact, the following tool can help.

Free ToolURL to Markdown ConverterConvert any public web page URL to Markdown. Preserves headings, tables, lists, and links — perfect for LLM and RAG preprocessing, research notes, and archiving web articles.Try it now →

FAQ

Q. What is auto mode in Claude Code?
It is a mode that runs without routine permission prompts. Instead of a person confirming each action, a separate AI called a classifier model reviews actions before they run. Anything that escalates beyond your request or targets unrecognized infrastructure is blocked.
Claude Code Official Documentation (Permission modes)
Auto mode lets Claude execute without routine permission prompts. A separate classifier model reviews actions before they run, blocking anything that escalates beyond your request, targets unrecognized infrastructure, or appears driven by hostile content Claude read. Claude Code Official Documentation (Permission modes)
Q. Is auto mode enabled by default?
On Pro, Max, and Team plans, auto mode is the built-in starting mode. That applies to terminal and VS Code extension sessions; claude -p, SDK usage, and the very first session after install start in Manual. It also requires v2.1.228 or later on macOS, Linux, and WSL and v2.1.233 or later on Windows; earlier versions default to Manual.
Claude Code Official Documentation (Permission modes)
On Pro, Max, and Team plans, auto mode is the built-in starting mode. / The built-in `auto` default requires Claude Code v2.1.228 or later on macOS, Linux, and WSL, and v2.1.233 or later on native Windows. On earlier versions, the built-in default is Manual. Claude Code Official Documentation (Permission modes)
Q. Does auto mode mean it is safe?
No. The documentation states plainly that it reduces permission prompts but does not guarantee safety. It is meant for tasks where you trust the general direction, not as a replacement for review on sensitive operations.
Claude Code Official Documentation (Permission modes)
Auto mode reduces permission prompts but does not guarantee safety. Use it for tasks where you trust the general direction, not as a replacement for review on sensitive operations. Claude Code Official Documentation (Permission modes)

Related Tools

Related Tool Categories

Articles