How do I turn auto mode off in Claude Code?

Auto mode is a permission mode you can leave at any time. Press Shift+Tab in the Claude Code CLI, or use the mode dropdown in the desktop app, to switch to another mode. A default you set for yourself is not overwritten by the August 14 change, and you only move to auto mode if you accept the one-time switch prompt.

Can my organization disable auto mode for everyone?

Yes. Administrators can set defaultMode in Claude Code managed settings to pin a different permission mode across the organization, or set disableAutoMode to remove auto mode entirely. A default that an organization manages is left alone by the August 14 rollout, so nothing changes for those developers until the admin changes it.

What does auto mode in Claude Code still block?

The auto mode classifier blocks actions that are irreversible, destructive, or pointed outside your environment, with unconditional hard deny rules for data exfiltration. When it blocks, Claude usually looks for a safer route or asks you directly. Your own permissions.ask rules still force a prompt, and permissions.deny rules stop an action before the classifier ever sees it.

How does auto mode compare to approving each prompt by hand?

Anthropic tested both. Human testers spotted dangerous commands 13.6% of the time, while auto mode caught 89% of them. Sessions ended in production-level harm 6.3% of the time under manual approval and 2.4% under auto mode. Anthropic frames the gap as the cost of reflexive clicking: 97% of prompts were approved anyway.

Does auto mode work on Bedrock, Google Cloud or Microsoft Foundry?

Auto mode itself is available on every provider, including the Anthropic API, Claude Platform on AWS, Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry. What differs is the default: on Enterprise, API and those cloud providers auto mode stays opt-in for now, and Anthropic plans to make it the default there within a month of the August 14 consumer rollout.

Anthropic · 2026-08-07 · major

Claude Code auto mode becomes the default — a classifier replaces most prompts

Anthropic is making auto mode the default permission mode in Claude Code for Pro, Max and Team plans on August 14, 2026. A safety classifier judges each tool call instead of asking you, and blocks irreversible or destructive actions.

Claude Code artwork from Anthropic's auto mode announcement

Claude Code stops asking before most tool calls and routes them through a safety classifier instead.

Key specs

Dangerous actions caught	89% (vs 13.6% by human testers)
Prompts people approved by hand	97%

Quick facts

Maker	Anthropic
What changed	Auto mode becomes the default permission mode
Default from	August 14, 2026
Plans first	Pro, Max and Team
Enterprise / API	Opt-in for now, default planned within a month
Configuration	autoMode block in user or managed settings

What is it?

Auto mode becomes the default permission mode in Claude Code on August 14, 2026, starting with the Pro, Max and Team plans. Instead of a permission prompt before each tool call, a safety classifier decides whether the call runs. Anthropic says people approved 97% of the prompts they were shown, so the prompt had stopped working as a real check. Anyone who already pinned their own default keeps it.

How does it work?

The classifier scores every tool call and blocks anything irreversible, destructive, or aimed outside your environment. Deny and explicit ask rules are read before the classifier, so they still block or force a prompt. On top of that sit hard deny rules for data exfiltration, a git state check before destructive commands, and prompt-injection screening. After three blocks in a row, or twenty across a session, Claude Code falls back to manual approval.

Why does it matter?

Dropping the prompt changes how long an agent can work unattended. Anthropic reports 9x longer stretches between interruptions, and about 25% more pull requests shipped by Team and Enterprise users running auto mode. Its own measurements found production-level harm in 2.4% of auto mode sessions against 6.3% of hand-approved ones, and no successful attack in its prompt-injection evaluation.

Who is it for?

Claude Code users on Pro, Max and Team plans

Frequently asked questions

How do I turn auto mode off in Claude Code?: Auto mode is a permission mode you can leave at any time. Press Shift+Tab in the Claude Code CLI, or use the mode dropdown in the desktop app, to switch to another mode. A default you set for yourself is not overwritten by the August 14 change, and you only move to auto mode if you accept the one-time switch prompt.
Can my organization disable auto mode for everyone?: Yes. Administrators can set defaultMode in Claude Code managed settings to pin a different permission mode across the organization, or set disableAutoMode to remove auto mode entirely. A default that an organization manages is left alone by the August 14 rollout, so nothing changes for those developers until the admin changes it.
What does auto mode in Claude Code still block?: The auto mode classifier blocks actions that are irreversible, destructive, or pointed outside your environment, with unconditional hard deny rules for data exfiltration. When it blocks, Claude usually looks for a safer route or asks you directly. Your own permissions.ask rules still force a prompt, and permissions.deny rules stop an action before the classifier ever sees it.
How does auto mode compare to approving each prompt by hand?: Anthropic tested both. Human testers spotted dangerous commands 13.6% of the time, while auto mode caught 89% of them. Sessions ended in production-level harm 6.3% of the time under manual approval and 2.4% under auto mode. Anthropic frames the gap as the cost of reflexive clicking: 97% of prompts were approved anyway.
Does auto mode work on Bedrock, Google Cloud or Microsoft Foundry?: Auto mode itself is available on every provider, including the Anthropic API, Claude Platform on AWS, Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry. What differs is the default: on Enterprise, API and those cloud providers auto mode stays opt-in for now, and Anthropic plans to make it the default there within a month of the August 14 consumer rollout.

Try it

Press Shift+Tab in the Claude Code CLI to switch permission modes, then run `claude auto-mode config` to print the rules in effect.