🧠 CRAFT: Collaborative Reasoning Agent for Flexible Teamwork

An intelligent, modular AI agent for Minecraft, powered by Mineflayer, LangChain, and natural language processing. The bot can interpret and execute user instructions written in natural language.

🚀 Features

• ✅ Natural language understanding and command execution
• 🧱 Mining, placing, crafting, and inventory management
• 🧭 Autonomous pathfinding and navigation
• ⚔️ Mob detection and combat capabilities
• 📦 Chest interaction (depositing and retrieving items)
• 📸 Screenshot capture and image-based reasoning via Puppeteer
• 🧠 Modular, tool-based reasoning using LangChain agents

📁 Project Structure

minecraft\_agent\_mineflayer/
├── agent/                # LangChain agent and tool integration
├── tools/                # Custom Minecraft tools
├── utils/                # Code utils
└──  index.ts              # Main entry point

🛠️ Requirements

• Node.js 18+
• A Minecraft server (tested on version 1.20)
• An OpenAI API key

📦 Installation

git clone https://github.com/benedettoscala/minecraft_agent_mineflayer.git
cd minecraft_agent_mineflayer
npm install

⚙️ Environment Variables

Create a .env file in the root directory and define the following:

OPENAI_API_KEY=your_openai_key
IP=your_server_ip
PORT=your_server_port
BOT_USERNAME=your_bot_username

🔧 Compile the TypeScript Code

tsc

▶️ Running the Bot

Start the bot with:

node build/index.js

💬 Example Commands

You can interact with the bot using natural language. Example prompts:

• bot go near the tree
• bot break that stone block
• bot attack the zombie
• bot open the chest and get the sword
• bot craft a wooden pickaxe

🧠 LangChain Agent Integration

The bot uses LangChain's agent framework to dynamically plan and execute tasks. Each Minecraft action (e.g., mining, following players, interacting with chests) is implemented as a LangChain tool with Zod validation schemas.

Based on the user's input, the agent selects the appropriate tool and executes the related Minecraft action.

🤝 Contributing

Contributions are welcome! Feel free to open an issue or submit a pull request. Please ensure that your code is clean, well-commented, and follows the existing structure.