AI for indie game developers, without giving up the project

Ask an indie developer why they will not use an AI game tool and the answer is rarely about the code being bad. It is that the project ends up in a proprietary format, on someone else server, in a tool that may be priced differently next year. That is a reasonable thing to refuse.

Flockbay is built the other way round. The engine is Godot, which is open source. Your project is an ordinary Godot project on your own disk. Flockbay writes GDScript into it. If you stopped using the app, you would still have a Godot project and Godot would still open it.

That is the part worth checking before anything else, because everything else is a preference and this one is structural.

It is worth having on the code nobody enjoys and everybody rewrites. Save systems. Inventory. Dialogue plumbing. A settings menu with working key rebinding. Object pooling. The state machine that every project needs and nobody wants to write for the fifth time.

It is not worth reaching for on the things that make your game yours. The feel of the movement, the pacing of the level, what the world looks like. Those are the reason to be an indie in the first place and they are not a prompt.

The split in this app is along that line on purpose. Flockbay writes code. The world gets built by hand, by you, in a builder that behaves like a level editor. There is no path where you describe your level and something invents it behind your back.

Plenty of people arriving at this search do write GDScript, and this page should be honest about the fit. The app is designed around someone who does not, which shows up in the interface: it wants you to describe and to judge rather than to edit.

But the project underneath is a normal Godot project, so there is nothing stopping you opening it, reading what was written, and changing it. If your objection to AI tools is that you cannot see the code, that objection does not apply here.

What it will not do is behave like an autocomplete in your editor. It is a builder that writes into your project, not a plugin that lives beside your cursor.

More on this: Is AI game making cheating?

The download is free and the world builder is free to use. Which builder you get is the first question a new project asks — 2D or 3D, two pictures, no default — and both halves are free. On the 3D side: landscape, weather, water, cameras, NPC behaviour, your own assets, play, playtesting and publishing. On the flat side: ground you paint as tiles, and sprites you place by hand out of a Catalogue that is 90.8% sprites, 55,073 of its 60,648 objects. Flockbay is free on your own AI: your Claude or ChatGPT plan, your own agent, or an API key. Flockbay builds from a library of ready-made objects, or with Godot's own shapes. There is no per-seat or enterprise tier, because it is a desktop app rather than a service.

Building requires an internet connection, since the AI and world data come over the network. The game you built runs locally without one, on the copy of Godot bundled with the app.

Windows both ship. The download is free to try before any of this matters to you.

More on this: An AI game maker with no sign up: Flockbay needs an account, and here is why

Writing code faster does not make a two-year project a two-month one. It moves the bottleneck, and the new bottleneck is you: playing, deciding, building the world, cutting things.

The failure this creates is specific. When systems are cheap, you add systems. A project with nine mechanics and no shape is easier to reach than it used to be, and it is exactly as unfinishable as it always was.

The useful discipline is unchanged. Decide what the game is, build the smallest version that contains it, play it, and only then add. The tool makes each attempt cheaper, which is worth a great deal, and it will not tell you when to stop.

In a project built this way, the hours move to playing and to the world. That is a different distribution from a hand-coded project and it takes a little adjusting to.

It is also the distribution most solo developers say they wanted. The complaint is rarely that there was not enough scripting; it is that there was never time for the level to be good.

Worth planning around. If code is no longer the constraint, the schedule should stop being built around it, and the extra time should go where the game is actually judged.