AI strategy game maker for people with a rule set in mind

People who want to make strategy games can usually describe them precisely. Three resources. Units cost two of them. Cavalry beats archers, archers beat spears, spears beat cavalry, and the ratio is not quite three to one so that the counter is a nudge rather than a rout. You can say that at a table and be understood.

That is unusual and it is an advantage. Most genres are specified by feel; strategy is specified by rules. And rules translate almost directly into code, which makes this the genre where writing the code for you costs the least in lost intent.

So the workflow fits. You explain the rules to Director 1, it writes GDScript into your Godot project, and you play a match to find out whether the rules you described produce the game you imagined. Usually they do not, the first three times, and that is the interesting part.

There is no mode switch to pick. Hex-based turns, simultaneous resolution, real-time with pause, card-driven activation: they are all just rules, and all reachable by describing them. What varies is how much content each demands from you.

Turn-based tends to be the kinder starting point because the systems are easier to reason about while playing, and because a turn-based prototype can be evaluated slowly. Real-time asks more of the map and of the interface before anything reads clearly.

Whatever the timing model, the map is where the design lands. You build it by hand in the builder: terrain, elevation, water, chokepoints, the resources that make one corner worth fighting over. Placing that directly is faster than describing it, and it lets you notice a problem while you are creating it.

The builder behaves like a level editor rather than a chat window. That is deliberate. The parts of a game where taste decides everything should be under your hands.

Every strategy developer eventually learns that a spreadsheet cannot tell you whether a unit is overpowered. It can only tell you the numbers are consistent. The verdict comes from playing, and preferably from someone else playing while you watch them ignore the unit you were proudest of.

Because the code is cheap to change, you can act on that immediately. That is the real benefit here: not that the code appears, but that changing it does not cost you a weekend and so you actually do it.

A desktop application for Mac and Windows. Downloaded, installed, with Godot bundled inside so the game runs locally on your machine. Your project is a normal Godot project.

Building needs an internet connection since the model, asset generation and world data all come over the network. Playing what you built does not. There is no browser version.

Free to download with a free plan. Indie is $10 a month, Pro is $50, and both buy more usage rather than a different engine.

Before the roster and before the map, decide how long a match should take. It sounds like a detail and it governs the whole design. A twenty minute match cannot support a deep tech tree, because nobody reaches the end of it. A three hour match needs mid-game structure or it sags.

Deciding it early also gives you a test you can run. Play a full match, look at the clock, and see whether the interesting decisions were spread across it or clustered in the first five minutes. Most first drafts front-load their decisions and then coast, and that is fixable once you can see it.

Fog of war is usually treated as a feature to switch on, but what it really does is decide how much of your game is guessing. Full information turns strategy into calculation. No information turns it into a coin toss. The interesting games sit somewhere in between and are deliberate about where.

It is worth asking what the player should be uncertain about. Enemy position, enemy composition, enemy intent: these are separate, and hiding a different one produces a completely different game from the same underlying rules.