Maintained by @uukelele-scratch, @Sweaterdog, @riqvip, @MrElmida, and the community.
Crafting minds for Minecraft with LLMs and mineflayer!
Note
This is a lightweight reimplementation of mindcraft, built from scratch designed to run on low-end hardware. It does not include most features present in the community edition or the original mindcraft.
The open-source platform for crafting intelligent, collaborative agents in Minecraft using Large Language Models.
mce-nano is a lightweight, open-source platform for running intelligent, LLM-powered Minecraft bots using mineflayer. It is designed for simplicity, low resource usage, and easy customization, making it ideal for running on low-end hardware, servers, or as a starting point for your own Minecraft AI experiments.
- Minimalist: No bloat, just the essentials for LLM-driven Minecraft automation.
- Multi-Provider: Supports OpenAI, Gemini, Andy API, Ollama, Pollinations, and more.
- Configurable: Each bot can have its own model, provider, and personality in config.yml.
- Conversation-Aware: Optionally saves and reloads conversation history per bot.
- Docker-Ready: Simple Dockerfile and docker-compose for easy deployment.
- Community-Driven: Maintained by contributors from the Mindcraft community.
Feature | mindcraft-ce (Community Edition) | mce-nano (Nano Edition) |
---|---|---|
Development Status | Active | Active |
Minecraft Version | Up to 1.21.6 | Up to 1.21.6 |
Node.js Version | v18+ (v22 recommended) | v18+ |
Setup Complexity | Complex (many dependencies) | Minimal (essential deps only) |
Multi-Bot Support | Yes | Yes (per-bot configs) |
Auto-Actions | Manual commands | Auto-eat, auto-defense, idle actions |
Chat Restrictions | Global only | Public/private + player whitelist |
Resource Usage | High (full feature set) | Ultra-lightweight |
Memory Footprint | Large codebase | Minimal core |
Docker Image Size | ~500MB+ | <200MB |
Plugin System | Complex plugin architecture | Simple, hackable core |
Learning Curve | Steep (many features) | Gentle (focused essentials) |
Fork Friendly | Heavy customization required | Easy to modify & extend |
Boot Time | Slower startup | Fast startup |
Model Providers | 15+ providers | 10+ essential providers |
Low-End Hardware | Struggles on weak systems | Optimized for Raspberry Pi+ |
Configuration | Multiple config files | Single config.yml |
Ideal For | Production & feature-rich bots | Learning, prototyping, servers |
- Minecraft Java Edition (up to v1.21.6)
- Node.js (v18 or newer)
- git
- (Optional) API key for your chosen LLM provider (see keys.example.json)
- (Optional) Docker for containerized deployment
-
Clone the repo:
git clone https://github.com/mindcraft-ce/mce-nano.git cd mce-nano
-
Install dependencies:
npm install
-
Copy and edit your keys:
- Copy keys.example.json to keys.json and add your API keys (see FAQ for help).
-
Edit your bot config:
- Edit config.yml to set up your bot(s), model, provider, and options.
-
Start Minecraft and open to LAN (default port: 55916).
-
Run the bot:
node main.js
Or use Docker:
docker-compose up --build
- Bots: Each entry in config.yml under
bots:
defines a bot with its own username, model, provider, and options. - Providers: Supported providers include OpenAI, Gemini, Andy API, Ollama, Pollinations, and more.
- Conversation: Enable
save_conversation: true
to persist chat history for each bot. - Interactive:
idle_timeout_seconds
andidle_message
can make the bot active when nothing is happening.
See the comments in config.yml for more details.
Bots support a wide range of commands, including movement, inventory, crafting, and more. Commands are listed in prompt.txt.
Example usage:
!goToPlayer("player_name", 2)
!searchForBlock("oak_log", 16)
Parameters can be quoted or unquoted, and the parser is flexible.
- See FAQ.md for solutions to common errors and setup questions.
- Join the Discord Support for help and discussion.
Maintained by @uukelele-scratch, @Sweaterdog, @riqvip, @MrElmida, and the community.
MIT License. See LICENSE for details.