Overview
Mindcraft integrates large language models with Mineflayer to create programmable, collaborative bots within Minecraft. It supports multiple model backends and provides profiles, task suites, and evaluation tools to benchmark agent behavior.
Key Features
- Multi-backend model support: OpenAI, Anthropic, Gemini, Replicate, Hugging Face, Ollama, and others.
- Task-driven evaluation: Task definitions and tooling to measure agent performance inside the game.
- Extensibility: Profile-based configuration, Docker support, and a variety of community tutorials and demos.
Use Cases
- Build automated bots and collaborative agents inside Minecraft for education, research, or entertainment.
- Use as a platform to evaluate embodied/virtual agent capabilities and compare LLM-driven strategies.
- Rapid prototyping of agent behaviors using provided profiles and configuration.
Technical Characteristics
- Primarily JavaScript/Node.js with supplementary Python components for task tooling.
- Profile-and-task-driven architecture (
profiles,andy.json) that enables reproducible agent setups. - Active community, academic citation (arXiv), and frequent releases.