Claude Code auto mode has become the default permission setting for users starting today.

Sessions for Pro, Max, and Team users now use it unless a user or administrator has pinned another setting.

According to Anthropic, auto mode replaces repeated approval prompts with a classifier that checks each tool call for irreversible, destructive, or out-of-bounds actions.

If it blocks something, Claude can try a safer route or ask permission. After repeated blocks, the session falls back to manual approval.

As part of the change, Anthropic says it stopped charging for “small number of extra tokens per tool call” used for the auto mode classifier last week.

Anthropic says its study of 1,053 paid testers found people caught 13.6% of dangerous commands, compared with 89% for auto mode. Human performance fell to about 5% after 50 prompts.

Still, Anthropic cautions that classifiers cannot eliminate risk and recommends human review for production changes.

The shift should make Claude Code’s long-running work more practical, building on routines, voice mode, and its in-app Mac browser. Anthropic says auto mode users at Team and Enterprise customers ship about 25% more pull requests.

While it may sound risky, this is probably the right default for experienced users. Still, users should establish some trust with coding agents before relying on auto mode.

It’s also interesting that Anthropic is taking this step. For comparison, OpenAI actually opted out of auto mode for its most powerful version of GPT-5.6 as an extra precaution at one point.