An AI game builder where you still do the building

Making a game is two very different kinds of work. One is arranging things until they look and feel right, which is judged by eye and cannot be specified in advance. The other is writing the code that makes them behave, which is exact, tedious and largely the same in every project.

Most AI game tools try to do both from a prompt. The result is that the part you have opinions about gets guessed at, and the guess is never quite it. You end up describing a world in sentences, which is slower and worse than placing it.

Stage Engine splits them. The world is yours and you build it with your hands in a builder that works like a level editor. The code is written for you by Director 1, as GDScript, inside your project.

You start with terrain. Raise it, carve it, put water where you want water. Add weather and a sky. Place buildings, props, characters. Move the camera around and see how it reads. Nothing here goes through a text box.

This is not a compromise or a fallback for when the AI cannot manage. It is faster. Dragging a hill into shape takes a second. Describing that hill precisely enough that something else could produce it takes a paragraph and still lands somewhere you did not mean.

The builder is also where you see whether the game is any good, because you can drop into the world and walk it whenever you want.

Everything that is code. Player movement, inventory, dialogue, enemy behaviour, objectives, saving, menus. You say what should happen and it writes the GDScript inside your project.

There is no third path where Director 1 quietly rearranges your world for you. Either you did it in the builder or you asked for code. That is deliberate, because a tool that changes your level without telling you is a tool you cannot trust with a level you care about.

You then play it and say whether it is right. That is the loop, and it is meant to run many times.

It is a desktop app for Mac and Windows, downloaded and installed. The engine underneath is Godot, open source. Your project is a real Godot project on your own disk.

The game runs locally on the copy of Godot that ships with the app. No streaming, no browser preview. Building needs an internet connection because Director 1, asset generation and world data come over the network. Playing what you built does not.

The download is free, there is a free plan, Indie is $10 a month and Pro is $50. Paid plans buy more usage rather than a different engine.

You do not need a design document to start. A useful first hour is building a small piece of world you like the look of, asking for a way to walk around in it, and seeing how it feels.

That is a better opening than planning, because what you learn is real. A world that feels good to move through will suggest a game. A game planned on paper often turns out not to want the world you imagined.

The project can grow outward from that. Add a reason to be there, then something to do, then something that can go wrong. Each step is a description and a play session, and each one tells you what the next should be.

You will play something, decide it is wrong, and ask for it again. That happens many times and it is not a sign that anything failed.

The more specific the complaint, the better the next version. It is too slow is less useful than it takes too long to get moving after I press the key, and the second is a sentence anyone can write without knowing what the code looks like.

This is the part of the process that is genuinely yours. You are the only person who has played your game, and the whole tool is arranged around getting your reaction back into the project quickly.