Epic provides a plugin for Unreal Engine 5 that lets game developers train AI-controlled characters and agents by having them learn through trial and error — similar to how a person gets better at a task by practicing. The tool, called Learning Agents, launched as an experimental feature with Unreal Engine 5.3 in September 2023 and is aimed at developers, not players.

Learning Agents supports two training approaches. Reinforcement learning rewards the AI agent for making good decisions — such as navigating a course correctly — and penalizes it for mistakes, allowing the agent to gradually improve. Imitation learning records a human player performing a task and trains the AI to mimic those behaviors. Training is done via Python-based scripts within the plugin. Use cases include NPC behaviors, physics-based character animation, and game-playing agents. The tool has been featured in Hugging Face's deep reinforcement learning course, indicating some adoption in the broader AI development community. As of 2025, Learning Agents remains marked as Experimental in the engine and has not been widely adopted in shipped commercial games.