Agentic coding

Agentic coding is an approach in which software components are designed as autonomous agents that perceive their environment, pursue goals, and interact with other agents. These agents make local decisions based on their internal state and rules, which can lead to emergent system-level behaviors. The paradigm is common in game development, simulations, multi-agent systems and AI, and is used to model independent actors such as NPCs, robots or simulated entities.

In a Unity game you might use agentic coding to make NPCs move and react realistically: each NPC is an agent with goals (e.g., patrol, flee, attack), a perception module, and a decision layer (such as a behavior tree or state machine). By allowing agents to communicate locally you can create group behaviors like cooperating enemies or a fleeing crowd without a central controller. Another example is a traffic simulation where each vehicle is an agent that makes choices based on surroundings and traffic rules, letting you study flows and congestion.