Runs LLM agents as Minecraft players through Mineflayer, with persistent sessions, per-agent shell sandboxes and multi-agent teams.
mindcraft-bots/mindcraft
Minecraft bots driven by LLMs, built on Mineflayer, where each AI agent plays the game and can chat with players.
- Stars
- 5,979
- Forks
- 987
- Open issues
- 63
- Last commit
- 11 days ago
What is it?
Mindcraft is a framework for running AI agents inside Minecraft Java Edition. It connects a large language model to a bot built on Mineflayer, so the agent can perceive the world, chat with players, and carry out in-game actions such as gathering resources and crafting items.
Agents are configured through JSON profiles that set the chat model, a separate code model, an optional vision model, embedding and voice settings, along with personality prompts and examples. The project supports many LLM providers, including OpenAI, Gemini, Anthropic, DeepSeek, and local models through Ollama. It can run against a LAN world on localhost or connect to online servers with a Microsoft account, and it includes a task runner for predefined goals.
Code generation is disabled by default because it lets the LLM write and execute code on your machine. The README recommends running it in Docker before connecting to any remote server.
Key features
- 01Connects LLMs to Mineflayer bots that play Minecraft Java Edition.
- 02Supports many model providers, including OpenAI, Gemini, Anthropic, DeepSeek, and Ollama.
- 03Lets you assign separate models for chat, coding, vision, and embeddings.
- 04Runs predefined tasks with inventory and target goals via a task JSON file.
- 05Can run multiple agents from separate profile files at once.
Use cases
- Researching multi-agent LLM collaboration in an embodied, game environment.
- Experimenting with how different LLMs behave as in-game characters.
- Building scripted or goal-driven Minecraft tasks to evaluate agent performance.
- Running local-model Minecraft agents through Ollama without external API calls.
Quick start
- 1
Install Node.js (v18 or v20 LTS recommended)
- 2
Download the latest release or clone the repository
- 3
Rename keys.example.json to keys.json and add an API key
- 4
Install dependencies from the project directory
npm install
Things to know
- Requires an API key from a supported LLM provider, and OpenAI is the default.
- Requires a Minecraft Java Edition installation, with support up to v1.21.11.
- Allows an LLM to write and run code on your machine when coding is enabled, so it must not be connected to public servers with that setting on.
In collections
Topics
FAQ
- What is mindcraft?
- Minecraft bots driven by LLMs, built on Mineflayer, where each AI agent plays the game and can chat with players. Mindcraft is a framework for running AI agents inside Minecraft Java Edition. It connects a large language model to a bot built on Mineflayer, so the agent can perceive the world, chat with players, and carry out in-game actions such as gathering resources and crafting items.
- Is mindcraft free to use? What license does it have?
- mindcraft is open source under the MIT license. Check the license terms before using it in commercial products.
- Is mindcraft still maintained?
- Actively maintained. The last commit to mindcraft was on Oct 27, 2026.
- What can I use mindcraft for?
- Researching multi-agent LLM collaboration in an embodied, game environment. Experimenting with how different LLMs behave as in-game characters. Building scripted or goal-driven Minecraft tasks to evaluate agent performance. Running local-model Minecraft agents through Ollama without external API calls.
- How do I install mindcraft?
- Install Node.js (v18 or v20 LTS recommended) Download the latest release or clone the repository Rename keys.example.json to keys.json and add an API key Install dependencies from the project directory: npm install
- Where can I get mindcraft?
- The source code and installation instructions are on GitHub: https://github.com/mindcraft-bots/mindcraft
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Unity ML-Agents Toolkit lets Unity games and simulations serve as environments for training AI agents with reinforcement and imitation learning.
A JavaScript API for building Minecraft bots that can join servers, move, dig, craft, chat and interact with the world.
Web-based, multi-user panel for deploying and managing Minecraft and Steam game servers across multiple machines, with Docker or process-based instances.
A 9B-parameter open-weight multimodal model for visual understanding, spatial reasoning, agentic tool use, and embodied-AI research, built on Qwen3.5 and C-RADIOv4-H.
A training framework for LLMs, VLMs, diffusion and embodied models on NVIDIA GPUs and Kunlun XPUs, built on Megatron-LM and torch-native backends.