Updated for Project2Markdown V2: This guide utilizes native --format=xml and automated minification flags to keep large Unreal Engine 5 projects strictly within LLM token budgets. For full architectural features, read the Project2Markdown overview.
Unreal Engine projects are largely composed of proprietary binaries: compiled shaders, geometry meshes, 4K textures, and .uasset packages that Large Language Models (LLMs) cannot parse directly. While Project2Markdown automatically detects and ignores binary files, core Blueprint logic remains embedded inside those binary assets.
To give an AI complete reasoning capability over your game architecture, you must first extract your node logic and level metadata as structured text. Follow this step-by-step pipeline to export, minify, and parse your UE5 project into clean AI context.
Phase 1: Setting Up a Clean & Isolated Workspace
Duplicate Your Project Folder
Create a duplicate copy of your entire Unreal Engine project. Working from a dedicated copy ensures your active production files and version control remain safe.
Clear the Content Folder in the Duplicate
Delete everything inside the Content directory of the duplicated project. This keeps the folder lightweight and prevents mixing binary assets with the text exports you are about to create.
Phase 2: Bulk Exporting Blueprints & Level Logic to Text (.t3d)
Open Your Original, Active Project
Launch your original project inside the Unreal Engine 5 Editor (not the duplicate), as the native bulk export action must be performed directly from the live editor.
Open the Content Browser Filter Panel
Navigate to the Content Browser and open the filter panel to isolate logic assets from 3D models and media.
Filter Exclusively for Logic Assets
Select Blueprints, Levels, and Materials. Keep heavy textures, audio, and static meshes unchecked to prevent unnecessary noise.
Select All & Execute Bulk Export
Press Ctrl+A to select all filtered logic assets, then right-click and choose Asset Actions > Bulk Export. This outputs full node connections, pins, and coordinates into text-based .t3d files.
Target the Duplicated Workspace
Set the export destination directory to the root of your duplicated project workspace.
Rename the Exported Folder to “Content”
Unreal Engine names the exported directory Game by default. Rename this folder to Content so the hierarchy matches your project structure exactly.
Phase 3: Generating Token-Optimized AI Context with P2M
Access Project2Markdown Releases
Visit the official GitHub repository and navigate to the Releases section to download the latest binary.
Download the Executable for Your OS
Download the standalone binary for Windows, macOS, or Linux. No background runtimes (Node.js or Python) are required.
Execute P2M with XML & Minification Flags
Place the executable inside the duplicated project root. Run the following command in terminal to remove empty lines, strip unnecessary comments, and structure the context for LLMs:
--format=xml adheres directly to Anthropic Claude’s context guidelines, and minification cuts down hundreds of thousands of auto-generated tokens from .t3d boilerplate.
Retrieve the Generated AI Context File
Project2Markdown processes the entire project tree concurrently across CPU cores, placing the final clean context file directly in your workspace within seconds.
Feed to Claude, ChatGPT, or Gemini & Restore
Upload the file to your AI model to trace Blueprint node bugs, analyze level hierarchy dependencies, or request automated refactoring. When the AI outputs new code, use p2m --restore to write it back to disk safely.
Frequently Asked Questions
Why can’t LLMs read Unreal Engine .uasset files directly?
.uasset files are compiled binary containers holding serialized byte data, shader binaries, and pointers. LLMs require human-readable text representations (like exported .t3d or C++ source code) to understand logic graphs.
What is the benefit of using XML format over standard Markdown for UE5?
Unreal Engine Blueprint exports contain complex parent-child node relationships. XML tags (e.g., <file path="...">) provide explicit hierarchy boundaries that advanced models like Claude 3.5 Sonnet parse with zero ambiguity.
Is Project2Markdown safe to run on enterprise game codebases?
Yes. P2M runs completely locally on your hardware with no internet access required for processing. Furthermore, its restore engine actively blocks path traversal attacks (../../) to prevent accidental directory overwrites.
Explore the Full Project2Markdown Ecosystem
Discover how P2M handles multi-threaded directory walking, BPE token estimation, and how it compares with tools like Repomix on massive codebases.
Read the Project2Markdown Architecture Guide →