UtilityAIKit: GameObject Utility AI for Unity


Give any agent organic, needs-based decisions with Utility AI on plain MonoBehaviours. Visual response-curve editor, live debug overlay, pure C#, no DOTS or dependencies.


by Tunnel Vision Studio


Price History +

Utility AI that feels at home in a normal Unity project.


UtilityAIKit lets your characters decide for themselves — eat when hungry, flee when threatened, rest when tired, wander when idle — by scoring their options instead of hard-coding them. No state machines to untangle, no behavior trees to rewire every time the design changes. Just needs, curves, and choices that emerge from them.


It is built the way most Unity games are actually made: plain MonoBehaviour and ScriptableObject. There is no DOTS or ECS to adopt, no baking step, and no package to install first — drop it in and it works on Unity 6 with the Built-in, URP, or HDRP pipeline.


How it works


- Considerations turn a raw signal (hunger, distance, energy, cover) into a 0–1 score through a response curve you draw. The curve is the tuning: a gentle ramp, a hard threshold, a falloff — each shapes how strongly an input pulls a choice.

- Decisions combine considerations into one score and carry the action to run when they win. Multiplicative scoring means any single deal-breaker can veto a choice; an optional compensation factor keeps decisions with many considerations fair.

- A UtilityAgent ticks through its decisions, picks the best above a threshold, and runs it — with optional inertia so agents commit to a choice instead of twitching between near-ties.

- A Blackboard holds each agent's state, and a reusable Brain asset lets one behavior drive a whole crowd, each with independent runtime state.


See the reasoning


Utility AI is only as good as your ability to tune it, so the kit ships the tools to watch it think:


- a visual response-curve window that shows every curve and, in Play mode, the live input→score point;

- a debug overlay with per-decision score bars and the winner highlighted;

- a Scene-view label floating each agent's current decision and score;

- a live readout right in the UtilityAgent inspector.


What's in the box


- Full, readable C# source — no black boxes, no compiled DLLs.

- A runnable demo scene: primitive creatures living out hunger, energy, threat, and wander behaviors with zero user input.

- EditMode and PlayMode tests.

- Code-first and asset-first workflows (build brains in code, or author them as Brain assets).


No strings attached


Pure C#. No cloud services, no analytics, no art or audio dependencies, nothing phoning home. The Input System and TextMeshPro are not required. One small runtime assembly with no external references.


Powerful decision making that stays simple, readable, and yours — for indie-friendly $24.99.