Voyager Minedojo is a self-learning AI agent that explores Minecraft, learns new skills, and makes discoveries without human guidance.

Voyager Minedojo is a self-learning explorer for Minecraft, powered by large language models such as GPT-4. It navigates the Minecraft world autonomously, acquiring new skills and making discoveries without human intervention. The system combines three core components: an automatic curriculum that guides exploration, a skill library that stores learned behaviors for reuse, and an iterative prompting mechanism that generates executable code to determine its next actions.
Voyager is notable for finding unique items, mastering the game’s tech tree, covering large map areas, and quickly learning new tasks in unfamiliar Minecraft worlds, outperforming previous methods across these benchmarks.
Voyager was created by Guanzhi Wang, Yuqi Xie, Yunfan Jiang, Ajay Mandlekar, Chaowei Xiao, Yuke Zhu, Linxi “Jim” Fan, and Anima Anandkumar — a collaboration involving NVIDIA, Caltech, UT Austin, Stanford, and ASU. It was introduced as the first embodied lifelong learning agent for Minecraft powered by an LLM, designed to continuously explore, learn skills, and make discoveries without human guidance.
Voyager is relevant for AI engineers, game developers, robotics engineers, NLP researchers, and AI researchers.
Voyager is an open research project. Learn more and access it through the official website at voyager.minedojo.org.
Related
Discover more tools in similar categories
Weekly picks of the best AI tools we've researched and added — no spam, unsubscribe anytime.
Join 25,000+ AI enthusiasts. No spam, unsubscribe anytime.