AI 3D game maker for people who want scale

The reason most AI game tools produce small flat scenes is not the AI. It is where the game runs. A hosted preview has to fit inside a tab and a hosting budget, and the honest consequence is short draw distances, few objects and simple light.

Flockbay puts the game on your machine. Godot ships inside the app, so pressing play launches your world natively with your GPU behind it. The ceiling is your hardware, and your hardware is a lot more generous than a tab.

That single decision is what makes terrain, weather, water and a populated world the ordinary case rather than the thing that makes the editor stutter.

Flockbay starts by asking questions, one round at a time, each with a recommendation, and writes a brief: what gets built first, what comes after, what is not in this version. Nothing is built until you approve it. Then it builds the world with its world tools: terrain sculpted and painted, water, sky, weather, plants, the player controller, characters and people, and props, all as nodes and resources in the scene tree.

It also writes the GDScript inside your project. That covers the camera, interaction, inventory, quests, enemy behaviour, save and load, and every unglamorous system that stands between a walkable scene and a game. Then it play-tests its own work: it runs the game, sends input events, reads stdout and the debugger, and captures the viewport to see what it built.

The AI is used at build time. What runs when you play is your game.

More on this: An AI game builder that builds the world, and leaves you the controls

You play, and you have opinions. The valley is too open. You cannot see the tower from the road and you should be able to. The night is too dark to navigate. All of those are worth acting on, and saying them is enough: Flockbay changes the world against the brief. There is nothing in that loop that requires you to read code, and nothing in it that removes your judgement about the result.

Some opinions are about one exact spot, and a sentence is a blunt tool for those. So the same world tools are buttons around the Godot editor's own 3D view: Sculpt, Paint, Plants, Place, Water, Sky, Weather, Player controller, NPCs, Camera. Lower the ridge until the tower shows, move the tree off the road, and Godot's undo takes it back if it was worse.

None of this requires you to learn a vocabulary first. You do not need to know what a navmesh is to say that the villagers walk through the wall, and knowing the word would not make your report any more useful.

More on this: AI city builder maker: simulation you can watch

It will not hand you a finished 3D game from a sentence: it agrees a brief first and builds the first version of it. It will not offer a menu of genre templates, because a template only works for a game that already exists.

And it will not pretend. If a thing did not work, it says what happened rather than producing something that looks like success. That is worth more to you than a demo that falls apart in the second session.

If your request depends on something you have not specified, it asks rather than inventing a plausible default, because a default you did not choose is a decision made on your behalf and hidden from you.

More on this: Who writes the code in Flockbay?

A desktop app for Windows, macOS and Linux, free to download, and every world tool is free. Flockbay is free, AI included: free hosted AI every day, or your own Claude or ChatGPT, your own agent, or an API key. Flockbay builds from a library of ready-made objects, or with Godot's own shapes.

Building needs an internet connection. The chat, asset generation and world data come over the network. Playing what you have built does not.

The common way a 3D project goes wrong here is not technical. It is a beautiful world that has nothing to do. Terrain is satisfying to ask for and to sculpt, the feedback is immediate, and it is very easy to spend three weeks on landscape and never ask what the player is supposed to want.

A useful discipline is to put the reason in early, even badly. Something to reach, something chasing you, something to collect. It can be a grey box on a hill. The point is that the moment there is a goal, you start judging the terrain by whether it makes the journey interesting rather than by whether it looks nice from the sky.

Almost every good design decision in an open world comes from that shift. A ridge is not there to look like a ridge, it is there so you cannot see what is behind it.

The brief pushes this way: Flockbay asks what the player does in the first minute before it raises a hill. But nothing will stop you making a landscape for its own sake, which is also a completely legitimate thing to do. It is just worth knowing which one you are doing.

More on this: An AI tower defense maker where the numbers are yours to set