Jargon:YOLO mode
Say it: YOH-loh mode · 1 min read
Say this at the party
“Letting an AI agent act without asking your permission first. Fast, handy, and risky anywhere it can do real damage.”
In plain English
Coding agents normally stop and ask before they run a command or change a file. That's safe, but you end up clicking "yes" all day.
YOLO mode switches the asking off. The agent just gets on with it, which is much faster and lets it keep working while you're away.
The name comes from "you only live once", which tells you all you need to know about the risk.
Where it came from
Cursor called its auto-run setting "YOLO mode" in late 2024, and the name stuck across the industry. Claude Code's version comes with a deliberately scary name: --dangerously-skip-permissions.
See it in action
You ask an agent to upgrade every library in your app. With YOLO mode off, it asks you 40 times. With it on, inside a sandbox, it's done before you've made dinner.
What people get wrong
- YOLO mode isn't reckless if you put it somewhere safe. Inside a sandbox with test logins, it's a sensible choice.
- Turning it on near your real accounts and files is the risky bit. An agent that's been tricked won't stop to ask.
Go deeper
- ArticleSimon Willison, 2025Designing agentic loopsThe case for YOLO mode, and how to do it safely.
- GuideAnthropic, Claude Code DocsConfigure permissionsThe settings between "ask me everything" and "ask me nothing".
- PostCursor Community Forum, 2024YOLO Mode is Amazing!An early user's excitement, risks and all.
Where it comes up
Related jargon
Drafted with AI, checked by the AI Council (Claude, ChatGPT and Gemini) and signed off by Phin.