GridWorld Agent Demo

This project simulates a simple virtual environment where an AI agent interacts with a 2D grid world using tools exposed via FastMCP. The agent can observe its surroundings, move around, pick up objects, and combine them using recipes to create new items — all governed by external LLM prompts via MCP.

Features

• 🧠 Agent with stateful inventory and position
• 🗺️ Grid-based world with emojis representing objects
• 🧰 Tooling via MCP for observe, move, pickup, drop, combine
• 📦 Recipe system for object interaction (e.g. making tea)
• ⚡ Streamlit UI for visualization
• 🔄 Auto-refresh on state changes

Getting Started

1. Install dependencies:

   pip install -r requirements.txt

2. Start the streamlit UI:

   streamlit run ui/app.py

3. Configure your MCP server. E.g. for Claude:

    "grid_world_server": {
        "command": "uv",
        "args": [
            "--directory",
            "C:\\path\\to\\project\\root\\",
            "run",
            "mcp_server.py"
        ]
    }