MCP Unreal Server

A server implementation for interacting with Unreal Engine instances through remote Python execution.

Features

• ? Unreal Instance Management
• Automatic discovery of Unreal nodes via multicast
• Real-time node status monitoring

• Resource listing through LSP-compatible clients

• ? Remote Execution

• Execute Python code in Unreal Engine environments
• Support for both attended and unattended execution modes

• File execution and statement evaluation modes

• ? Logging & Monitoring

• Detailed logging to file (mcp_unreal.log)
• Console logging with different verbosity levels
• Node connection health monitoring

Installation

# Clone repository
git clone https://github.com/your-org/mcp-unreal-server.git
cd mcp-unreal-server

# Install dependencies
pip install -r requirements.txt

Configuration

Network Settings

Configure multicast parameters in RemoteExecutionConfig:

# Default multicast settings (modify in server.py)
config.multicast_group_endpoint = ("239.0.0.1", 6766)

Logging

Modify logging configuration in server.py:

# Adjust log levels
file_handler.setLevel(logging.DEBUG)  # File logging
console_handler.setLevel(logging.INFO)  # Console logging

Usage

Starting the Server

python -m src.mcp_server_unreal.server

Supported Tools

1. Connect to Unreal Instance

   {
     "host": "239.0.0.1",
     "port": 6766
   }

2. Execute Python Code

   {
     "node_id": "<unreal-node-id>",
     "code": "print('Hello Unreal')",
     "unattended": true
   }

API Documentation

Resource Format

types.Resource(
    uri="unreal://<node_id>",
    name=f"Unreal Instance: {node_id}",
    description="Unreal Engine instance",
    mimeType="application/x-unreal"
)

Execution Modes

Troubleshooting

Common Issues: - No nodes discovered: Verify Unreal instances are running with MCP plugin - Execution timeout: Check firewall settings for multicast traffic - Connection drops: Monitor mcp_unreal.log for node status changes

License

Apache-2.0 License