AI agent glossary
Permission modes
Settings that control how much an agent may do without asking, from approving every action to editing files freely to running fully unattended.
Every agent that can change things faces a tradeoff between safety and speed. Permission modes let you choose where to sit on that spectrum for a given session instead of answering the same approval prompt over and over.
Common levels include asking before each file edit or command, automatically accepting file edits while still asking before shell commands, a read-only planning mode, and a mode that skips all prompts. Claude Code, OpenAI Codex, and other coding agents each offer their own version of these levels, usually combined with allow and deny lists for specific commands or tools.
Match the mode to the risk and the environment. Approve each step when you are learning a tool or working near production data. Reserve the skip-everything modes for isolated sandboxes or containers where the worst case is a deleted throwaway environment.
Learn more in the course
Related terms
Go from definitions to doing
Learn to use and build AI agents in 65 practical lessons.