Voyager Minecraft AI

Voyager is an open-source research project from MineDojo that introduces the first large language model (LLM)-powered embodied lifelong learning agent designed for the game Minecraft. The agent continuously explores the Minecraft world, acquires diverse skills, and makes novel discoveries without any human intervention. It leverages GPT-4 through blackbox API queries to drive its decision-making and skill acquisition, bypassing the need for model fine-tuning.

The system consists of three core components: an automatic curriculum that maximizes exploration, an ever-growing skill library of executable code for storing and retrieving complex behaviors, and an iterative prompting mechanism that incorporates environment feedback, execution errors, and self-verification to improve programs. Skills developed by Voyager are temporally extended, interpretable, and compositional, enabling rapid compounding of abilities and alleviating catastrophic forgetting.

Empirically, Voyager achieves significant improvements over prior state-of-the-art methods: it obtains 3.3× more unique items, travels 2.3× longer distances, and unlocks key tech tree milestones up to 15.3× faster. The learned skill library can be applied in new Minecraft worlds to solve novel tasks from scratch, demonstrating strong generalization.

Voyager is available as a GitHub repository under the MIT License. It requires Python ≥ 3.9 and Node.js ≥ 16 for installation. The project is primarily aimed at researchers and developers interested in embodied AI, reinforcement learning, and LLM-based autonomous agents. It is not a commercial SaaS product.