AI Modes
Pick the mode that fits the job, from a quick fix to a large system.
Every request you send runs in a mode. The mode decides how the AI works: whether it plans first, makes changes, or only reads. Picking the right one saves tokens and avoids changes you didn't want.
Which Mode Should I Use?
| You want to… | Use |
|---|---|
| Build a large system that touches many others | Super Agent Mode |
| Build, add or change something (most requests) | Agent Mode |
| Understand code or a system, without changing anything | Ask Mode |
| Find out why one specific thing is broken | Debugging Mode |
Not sure? Stay on Agent Mode. It's the default.
Where to Find It
Click Mode: under the chat box, next to your token count, and pick a mode before you send your request.
Agent Mode
In the app: "Normal agentic mode for general tasks and simple systems."
Use it for most requests: adding a system, changing how something works, fixing a bug you already understand. "Agentic" means it doesn't just answer. It looks at your project's scripts and folders, works through the files it needs, and makes the changes. Superbullet then syncs them into Studio through .
It still works one request at a time, on what you asked for. You check the result and ask for the next change.
Super Agent Mode
In the app: "Creates a detailed architectural plan for your review before building complex systems."
Use it when a request is big enough that a wrong start would be expensive: a system that touches several others, like a trading system that works with inventory, currency and data saving. Super Agent Mode writes a detailed plan first, for the structure of the system, and you review it before anything is built.
Read the plan properly. It's the cheapest point to catch a wrong idea, before tokens go into building it.
Ask Mode
In the app: "Read-only mode for debugging and exploration."
Ask Mode reads your project and answers. It doesn't change anything. Use it to understand before you change:
- how an installed library works, before you customize it,
- how a system you didn't write fits together,
- where something lives, like which script saves player data,
- why something might be behaving the way it is.
When you know what to change, switch to Agent Mode and ask for it.
Debugging Mode
In the app: "Surgical debugging with minimal, targeted debug prints."
Use it when one specific thing is broken and you don't know why. Instead of rewriting code, it adds a few targeted debug prints to narrow down the cause, keeping its changes to a minimum. Play the game in Studio to trigger the bug, and the prints show up in Studio's Output window.
Describe the bug precisely: what you did, what you expected, and what happened instead.
Debugging Mode or Agent Mode? If you already know what's wrong ("the jump pad launches players too high"), ask Agent Mode to fix it. If you don't know where the problem is, start with Debugging Mode.
It helps you track down the cause. In some cases it removes its debug prints afterwards; if any are left once the bug is fixed, ask it to take them out.
Modes and Models
The mode is how the AI works. The model, picked in AI Model above the chat, is which AI does the work. They're separate choices. See Best Practices for choosing a model.