ChiR24/Unreal_mcp

What it solves

This project provides a bridge between AI assistants (like Claude, Cursor, or VS Code) and the Unreal Engine editor. It allows an AI to directly interact with the editor to build levels, create Blueprints, design UI, set up materials, and manage gameplay systems without the user having to manually perform every action in the engine.

How it works

The system consists of a native C++ editor plugin called the MCP Automation Bridge that runs inside Unreal Engine. AI clients connect to this plugin using the Model Context Protocol (MCP) via two possible routes:

• Route A (Native HTTP): The client connects directly to the plugin's HTTP server (default port 3000) using a capability token for authentication.
• Route B (stdio): The client connects through a Node.js server (unreal-engine-mcp-server) which acts as a proxy via stdio and WebSockets (default port 8090).

Instead of exposing hundreds of individual tools, the plugin exposes a single unreal tool. The AI assistant uses a gateway flow: it searches for a capability (e.g., "spawn actor"), describes the specific action to get the required parameters, and then executes the action.

Who it’s for

Game developers and technical artists using Unreal Engine 5.0 to 5.8 who want to automate editor tasks or use AI to accelerate the level design, blueprinting, and world-building process.

Highlights

• Extensive Capabilities: Nearly 400 editor actions across levels, actors, Blueprints, UI, UMG, materials, Niagara effects, and AI behavior trees.
• Flexible Connectivity: Supports both direct HTTP and stdio transports to accommodate different AI clients.
• Native Integration: Built as a C++ plugin to ensure deep access to editor APIs on the game thread.
• Safety Mechanisms: Destructive actions require a specific consent grant to prevent accidental deletions.
• AI-Friendly Discovery: Uses a search-based system so the AI doesn't need to memorize hundreds of tool names.