Goals meet circumstances.
Combine beliefs, goals, action planning, and utility scoring. Attributes, relationships, and perceived signals give each agent a basis for choosing what to do next.
TOOLS / WARWEAVER SDK
NPCs with goals, memories, and reasons for what they do. Warweaver is our deterministic cognition runtime for developers building systemic game worlds.
Closed technical preview · Unity 6+ · Access by evaluation
A SYSTEMIC GAMEPLAY EXAMPLE
A scout spots danger. A guard might advance, a villager might flee, and a healer might stay to help. The same signal reaches agents with different goals and priorities. You define the available actions and rules; their decisions emerge from those systems.
An illustrative use case, not a running simulation.
Combine beliefs, goals, action planning, and utility scoring. Attributes, relationships, and perceived signals give each agent a basis for choosing what to do next.
Deterministic execution makes the same seed and command sequence reproducible. Reasoning traces help you inspect the decisions behind an unexpected outcome.
The kernel handles cognition. Your game handles movement, animation, physics, and rendering. Signals and action results connect the two through the Unity adapter.
INTEGRATION / UNITY 6+
Add WarweaverManager to a scene, configure your schema before the simulation starts, and connect your game to the adapter. Simulation ticks are separate from rendering.
Declare attributes, operators, and agent archetypes before the first tick.
Inject world events and perception. The kernel processes them on its simulation ticks.
Read intents, perform actions in your game, and report success, failure, or a blocked path.
Read telemetry and replay a recorded sequence to investigate an agent’s decisions.
FROM THE SDK REFERENCE
Define a danger signal during setup, then inject it when your scout detects an enemy. Its radius and duration determine where and how long it can be perceived. Each agent still needs goals and actions that tell it how to respond.
Integration excerpt. Requires an initialized adapter and a configured agent schema; it is not a standalone example.
C# / UNITY ADAPTER
// During schema setup, before simulation starts:
int dangerSignalId = _adapter.CreateSignal("Danger");
// When your game detects an enemy:
_adapter.InjectSignal(
signalId: dangerSignalId,
position: enemyPosition,
radius: 100f,
duration: 120 // simulation ticks
);
The kernel evaluates perception, goals, and plans. Your Unity code polls the emitted intents, executes movement or actions, and reports their results. It also supplies the world information the simulation needs. Warweaver does not replace your game’s physics, navigation, animation, or rendering systems.
The documented kernel uses fixed-point math, seeded random numbers, a canonical iteration order, and per-tick state hashing. Reproducing behavior requires the same simulation configuration, seed, and input sequence. Reasoning traces expose decisions at specific ticks; they do not automatically replay all of your Unity game state.
Start with a representative slice of your game: the actual agent behaviors, world signals, population, tick rate, and target hardware you need. Measure simulation cost, memory use, integration effort, and whether the resulting behavior serves your game. Population alone is not a performance guarantee.
CLOSED TECHNICAL PREVIEW
We’re working with teams exploring colony simulations, sandbox RPGs, and other systemic worlds. Evaluation starts with a technical conversation, followed by private SDK access for suitable projects and a review of the integration together.
The current preview supports Unity 6+. Our studio’s Godot prototypes are separate projects; a Godot SDK integration is not part of this preview.
The evaluation request form opens on warweaver.io.
Read the evaluation license games@warweaver.io