ReinforceMC

Training runs inside the game. A user defines custom rewards for in-game actions such as killing players or mobs, taking damage, healing, breaking and placing blocks, and crafting items, then starts, stops, pauses or resumes a training session from the mod's controls. Trained models can be saved and loaded again later, and an emergency stop halts training immediately if needed.

The mod ships with a GUI made of several configuration screens: a main configuration menu, a reward system setup, an input configuration, a training control panel, a performance monitor and template management. Reward tracking is shown in real time. An integration with DeepSeek lets a user message a DeepSeek Telegram bot, which can carry out actions in Minecraft through the Baritone pathfinding mod.

The mod requires computational resources, and it is intended for users with at least basic knowledge of machine learning. Training times vary greatly with hardware, network complexity and the learning task. ReinforceMC is published for Forge only, for 1.16.5, and the mod is released under the MIT License without being open source.