Claude's auto mode is moving from the coding tool into everyday apps. Anthropic announced on October 22, 2026, through Claude's official account that, starting next week, sessions across Claude web, desktop and mobile apps will run in auto mode by default on Pro, Max and Team plans. Instead of asking users to approve each step, an automated check blocks risky actions, and users can switch back to manual mode at any time.
Auto mode already had a run in Claude Code
The mechanism first landed in Claude Code this August: from August 14, auto mode became the default permission mode in Claude Code on Pro, Max and Team plans. It uses a separate classifier to review shell commands and tool actions. In Anthropic's published testing at the time, the classifier caught 89% of dangerous commands, while manual approval caught only 14% — people click too fast, and by the fiftieth prompt approval becomes reflex. This week's change carries the same idea from coding into ordinary sessions: when Claude runs multi-step tasks and calls tools inside the apps, the default gatekeeper shifts from you approving every step to an automated check reviewing actions first.
When the default changes, so does the trust model
What matters is not the feature itself but that it becomes the default. Most users keep whatever default they are given, so approval fatigue disappears — but judgment moves to the classifier with it. A system that catches 89% still lets roughly one in ten risky actions through, and Anthropic itself cautioned in the Claude Code era that auto mode reduces prompts without guaranteeing safety. Read alongside 1Password for Agents, which lets agents sign in without ever seeing your password, the industry direction is consistent: agents get broader permissions, while safety shifts from approving everything up front to automatic checks at runtime.
The practical advice for Pro, Max and Team users is simple: after next week's update, check which mode your sessions start in, and let auto mode handle routine work. For irreversible actions such as deletions, payments or sending messages on your behalf, switching back to manual approval remains the safer choice.
