YOLO Mode
YOLO mode is a setting that lets an AI coding agent edit files, run shell commands and call tools without asking for approval first. The name comes from "you only live once": you trade the safety of reviewing each action for an agent that never stops to wait for you.
What YOLO Mode Means
With YOLO mode off, a coding agent pauses before each consequential tool call and waits for a person to allow or deny it. With YOLO mode on, those prompts are skipped and the agent works through the whole task on its own. In some agents YOLO mode also switches off the sandbox that limits which files and network addresses commands can reach.
Most agents also offer a halfway setting that approves file edits automatically but still asks before running commands.
YOLO Mode Flags by Agent
| Agent | How to turn it on |
|---|---|
| Claude Code | claude --dangerously-skip-permissions |
| OpenAI Codex CLI | codex --yolo |
| Google Antigravity | Turbo preset in the IDE, agy --dangerously-skip-permissions in the CLI |
| Gemini CLI | gemini --yolo, or Ctrl+Y in a session |
| GitHub Copilot | copilot --yolo in the CLI, Allow all in VS Code |
| Cline | YOLO mode checkbox in settings, cline --yolo |
| OpenCode | "permission": "allow" in opencode.json, or opencode --auto |
The guide to YOLO mode in every coding agent covers each agent's config file, halfway settings and sandbox behavior.
Using YOLO Mode Safely
- → Run it in a container, VM or disposable checkout with no production credentials.
- → Commit before the agent starts, so every change can be rolled back.
- → Give it one scoped task, not an open-ended goal.
- → Prefer per-task YOLO over a session-wide switch.
YOLO Mode in AgentRQ
In AgentRQ, YOLO mode is set per task, not per session. Tasks with it off send each permission request to you on the web, your phone or Slack to allow or deny. A task with it on has every tool call approved automatically, and each call is still recorded in the tool call history.