MineDojo/MineDojo

Health v2 · maintenance only

Adoption not part of the score

2251 stars · 197 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

MineDojo is an AI research framework built on Minecraft for training open-ended, generally capable embodied agents. It provides a simulation suite with thousands of diverse tasks plus an internet-scale knowledge base of YouTube videos, wiki pages, and Reddit posts.

Use cases

• train embodied agents in minecraft
• benchmark reinforcement learning agents on diverse 3d tasks
• learn agent policies from youtube videos
• research open-ended generally capable ai agents
• use internet-scale knowledge base for agent training
• evaluate multimodal agents with mineclip rewards

When to choose

• you need a rich 3d simulation environment for embodied AI research
• you want to train agents with internet-scale human knowledge
• you need a standardized benchmark suite for open-ended agent learning

When to avoid

• you need a lightweight or fast-turnaround RL environment
• you don't want a Minecraft/Java dependency
• you need production agent deployment rather than research

Facets

framework · maturity maintenance

simulation machine-learning reinforcement-learning agent-framework benchmarking rag artificial-intelligence reinforcement-learning simulation python minecraft embodied-agents open-ended-learning benchmark-suite youtube-dataset mineclip neurips-award game-development research linux macos gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="MineDojo/MineDojo")