Anthropic is making Claude Code more autonomous; Auto Mode coming as default for paid users
In Auto Mode, Claude Code uses a classifier to review tool calls that determine whether the action is safe.
- Aug 10, 2026,
- Updated Aug 10, 2026 1:19 PM IST
Anthropic’s Claude Code is getting more autonomous starting August 14, as Auto Mode will become the default for Pro, Max, and Team users. With Auto Mode as default, Claude can execute coding tasks more independently; this means the AI will not ask for permissions or prompts repeatedly.
On the mentioned date, you may get a “one-time prompt” if you want to switch to auto mode. However, if you’ve already pinned auto mode as default, you may not see any major change.
Must read: AI assistant hacks gym booking system in first known Australian autonomous cyberattack
Claude Code to get Auto Mode as default
Anthropic shared a blog post announcing Auto mode as default in Claude Code, moving away from the manual permission-prompt system. In addition, Auto mode uses a small number of additional tokens for each tool call because of its classifier. Now, Anthropic says it will no longer charge Pro, Max, and Team users for those additional classifier tokens.
In Auto Mode, Claude Code uses a classifier to review tool calls that determine whether the action is safe. The system blocks potentially destructive, irreversible, or external actions and try to find a safer way to accomplish the task. If it can’t find a safer alternative, Claude can ask the user for permission before proceeding.
Anthropic claims that Claude Code in Auto Mode could perform large, complex tasks for hours without interruption. Among Teams and Enterprise users, people using Auto Mode reportedly ship around 25% more pull requests. It also highlighted that humans caught dangerous commands only 13.6% of the time, whereas Auto Mode's classifier caught 89% of dangerous commands.
Anthropic also found ‘approval fatigue” among users. Reportedly, people clicked “approve” on 97% of permission prompts, highlighting thatusers often approve actions without carefully checking them in long sessions. In contrast, the classifier's performance remained consistent.
Corporate customers, including Adobe, Nuro, Gusto and Garner Health, are already using Auto Mode as their default production setup. However, Anthropic also says that the Auto Mode does not fully eliminate, but reduces, the risks. Therefore, it recommends manual review for high-stakes changes to production systems.
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine
Anthropic’s Claude Code is getting more autonomous starting August 14, as Auto Mode will become the default for Pro, Max, and Team users. With Auto Mode as default, Claude can execute coding tasks more independently; this means the AI will not ask for permissions or prompts repeatedly.
On the mentioned date, you may get a “one-time prompt” if you want to switch to auto mode. However, if you’ve already pinned auto mode as default, you may not see any major change.
Must read: AI assistant hacks gym booking system in first known Australian autonomous cyberattack
Claude Code to get Auto Mode as default
Anthropic shared a blog post announcing Auto mode as default in Claude Code, moving away from the manual permission-prompt system. In addition, Auto mode uses a small number of additional tokens for each tool call because of its classifier. Now, Anthropic says it will no longer charge Pro, Max, and Team users for those additional classifier tokens.
In Auto Mode, Claude Code uses a classifier to review tool calls that determine whether the action is safe. The system blocks potentially destructive, irreversible, or external actions and try to find a safer way to accomplish the task. If it can’t find a safer alternative, Claude can ask the user for permission before proceeding.
Anthropic claims that Claude Code in Auto Mode could perform large, complex tasks for hours without interruption. Among Teams and Enterprise users, people using Auto Mode reportedly ship around 25% more pull requests. It also highlighted that humans caught dangerous commands only 13.6% of the time, whereas Auto Mode's classifier caught 89% of dangerous commands.
Anthropic also found ‘approval fatigue” among users. Reportedly, people clicked “approve” on 97% of permission prompts, highlighting thatusers often approve actions without carefully checking them in long sessions. In contrast, the classifier's performance remained consistent.
Corporate customers, including Adobe, Nuro, Gusto and Garner Health, are already using Auto Mode as their default production setup. However, Anthropic also says that the Auto Mode does not fully eliminate, but reduces, the risks. Therefore, it recommends manual review for high-stakes changes to production systems.
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine