Pat Simmons documents the full setup for connecting an AI coding agent to Unreal Engine through MCP. He installs the editor plugin, enables the required tool sets and troubleshoots a local port conflict, then gives the agent a reusable instruction harness that records setup steps and quality-assurance checks.
The first build uses Unreal's City Sample assets. Simmons asks Claude to assemble a realistic futuristic city, lets it work for about 45 minutes and then inspects the result as a playable character. The generated environment does not match the requested Dubai style, but it includes varied buildings, traffic and lighting, which makes the demo useful as both a capability test and a record of current limitations.
He next connects Cesium and Google Maps Platform data to stream a 3D model of New York City, then uses headless Blender scripts to create a custom neo-Gothic building for Unreal. The walkthrough shows that agents can coordinate editor configuration, asset placement and iterative fixes, while still requiring substantial downloads, credentials, restarts and human review. The creator's bootcamp and newsletter promotions are omitted.
Watch on YouTube
