Live class · Intro · 2 hours · Past session
Intro to PCG & AI
with Alex Coulombe
Wed, Sep 9 · 11am–1pm ET
Taught live on Zoom and recorded.
About this session
Procedural Content Generation from scratch, plus the part nobody else teaches yet: wiring Unreal to an AI agent over MCP and letting it author PCG graphs for you.
What we covered
This session explored Unreal Engine's Procedural Content Generation (PCG) framework and its integration with external AI tools using the Model Context Protocol (MCP). Alex Coulombe demonstrated how to configure the MCP server to let AI assistants create PCG graphs and assemble environments inside the editor. He then covered PCG fundamentals by building procedural graphs from scratch to scatter assets, randomize transforms, and carve out paths with splines.
- 01Connecting Unreal Engine 5.8 to external AI assistants using the UnrealMCP plugin and local server configuration.
- 02Prompting Codex via MCP to autonomously author a functional PCG graph with custom assets and materials.
- 03Structuring agent markdown files and markdown knowledge bases to maintain workflow memory across AI sessions.
- 04Analyzing AI-generated PCG graph networks, debugging coordinate space issues, and cycling deterministic seed values.
- 05Setting up a PCG Volume, sampling landscape data, and configuring debug visualization modes in the viewport.
- 06Spawning trees and foliage using the Static Mesh Spawner and Hierarchical Instanced Static Meshes.
- 07Randomizing scale, rotation, and offsets with the Transform Points node to correct slope alignment.
- 08Branching graphs to scatter multiple weighted assets, including rocks and bushes, across distinct layers.
- 09Clearing procedural foliage along a spline path using spline sampling, Runtime Virtual Textures, and Difference nodes.
AI summary of the session recording's transcript — the recording itself is the source of truth.
