This project provides an unofficial Unreal Engine plugin to control the engine via AI agents using a Machine Control Protocol (MCP). It targets AI researchers and developers seeking programmatic control over Unreal Engine environments for tasks like scene manipulation and asset creation, enabling AI-driven game development and simulation.

How It Works

UnrealMCP implements a TCP server within Unreal Engine, exposing a JSON-based command protocol. External AI systems, such as Claude for Desktop, can connect to this server to send commands for scene manipulation, asset creation, and Python script execution. This approach allows for direct, programmatic interaction with the Unreal Engine, abstracting complex API calls into a simpler, AI-friendly interface.

Quick Start & Requirements

• Install: Clone the repository into your Unreal project's Plugins directory. Regenerate project files, build the project, and enable the plugin within Unreal Engine.
• Prerequisites: Unreal Engine 5.5 (tested), C++ development environment for Unreal, Python 3.7+ for client scripts. Requires an AI model like Claude for Desktop.
• Setup: Configuration involves modifying claude_desktop_config.json to point to the plugin's run_unreal_mcp.bat script.
• Docs: modelcontextprotocol.io/quickstart/user

Highlighted Details

• TCP server implementation for remote control.
• JSON-based command protocol for AI integration.
• Editor UI integration for MCP server management.
• Client-side Python scripts for interaction.
• Supports basic operations like creating/modifying objects and executing Python scripts.

Maintenance & Community

• Project is marked as "VERY WIP" with plans for user extensions and more tools.
• A Discord server is available for support: discord.gg/abRftdSe
• Creator: kvick (@kvickart on X).

Licensing & Compatibility

• MIT License. Permissive for commercial use and closed-source linking.

Limitations & Caveats

The plugin is explicitly marked as "VERY WIP" and primarily tested on Windows. AI model compatibility beyond Claude for Desktop is experimental, and the author notes potential errors with Unreal Python scripting due to limited examples. Direct AI modification of projects carries inherent risks, necessitating strict source control and backups.