liveBook · Manning

Auto permission mode

As explained in Claude at Work

Auto permission mode is a Claude Code mode that uses an AI classifier to evaluate tool permissions instead of repeatedly requiring direct human approval.

Auto permission mode replaces the human approver with an AI classifier that evaluates tool permission requests. It can reduce approval interruptions, but it also delegates permission decisions—and some workflow choices—to the model. section 5.6

Choose a permission mode to match the work and its risks:

  • Use auto when you want fewer approval checks but still want a permission mechanism. It is not the same as bypassPermissions, which skips checks. section 5.6
  • Review the work afterward: an auto-mode scaffolding session created a feature branch, committed changes, and opened a pull request. section 5.7

Scaffold a walking skeleton

Start an auto-mode session and ask Claude Code to scaffold the components of a distributed application. Specify the intended scope, such as a walking skeleton, and check the result: Claude may inspect project files and make workflow decisions along the way. section 5.7

/exit

claude --permission-mode auto

Start a session in auto mode with this command. section 5.7

How Claude at Work covers it

How deep it goes
Defines the classifier’s role and demonstrates a long scaffolding session, but does not explain the classifier’s internals.
Examples
Scaffolds a distributed application’s walking skeleton and shows a session that creates a branch, commits changes, and opens a pull request.
What you can do after
Operate Claude Code in auto mode with `claude --permission-mode auto` and review the resulting changes and workflow decisions.
How it connects
Compares auto with manual, acceptEdits, plan, and bypassPermissions permission modes.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Claude at Work. The section links point to the parts of the book it draws on.