Unity offers the ML-Agents Toolkit, a free open-source tool that lets game developers and researchers train AI-controlled characters and agents in Unity game environments using machine learning techniques such as reinforcement learning and imitation learning.

Unity ML-Agents is an open-source project that enables games and simulations to function as training environments for intelligent agents using reinforcement learning, imitation learning, and neuroevolution via a Python API. Game developers can use it to create non-player characters (NPCs) that learn adaptive behaviors, automate game testing, or evaluate design decisions. Trained models are embedded back into the Unity runtime (now via the Inference Engine) for real-time execution on player devices. ML-Agents supports single-agent, multi-agent cooperative, and multi-agent competitive scenarios. It has been available since 2017 and integrates with the Unity Editor.