How It Works

The plugin provides a unified interface for interacting with multiple LLM APIs, including OpenAI, Anthropic Claude, XAI Grok, and Deepseek. It supports direct API calls for chat completions and structured outputs, as well as the Model Control Protocol (MCP) for deeper integration with tools like the Claude Desktop App and Cursor IDE. This allows AI models to directly generate and manipulate Unreal Engine assets, blueprints, and scenes via Python scripting.

Quick Start & Requirements

• Installation: Add the repository as a Git submodule or await Unreal Marketplace release. Enable the "Generative AI Support Plugin" in Unreal Editor. For C++ projects, include "GenerativeAISupport" in Build.cs.
• Prerequisites: Unreal Engine 5.1 or higher. For MCP features, install Claude Desktop App or Cursor IDE, and mcp[cli] via pip.
• Setup: Setting API keys involves environment variables (e.g., PS_OPENAIAPIKEY). Restarting the editor and IDE is required after setting variables.
• Documentation: Plugin Documentation 🛠️🤝, Plugin Example Project 🛠️.

Highlighted Details

• Supports multiple LLM APIs: OpenAI (GPT-4o, GPT-4o-mini), Anthropic (Claude 3.5 Sonnet/Haiku, Claude 4 Opus), Deepseek (R1), XAI (Grok 3).
• Model Control Protocol (MCP) enables AI-driven blueprint and scene generation, object spawning, and transformation.
• Offers C++ and Blueprint examples for API integration and MCP usage.
• Includes features like text streaming, structured outputs, and multimodal API support (some marked as "pro" or "available in pro").

Maintenance & Community

• The project is under rapid development with a warning against production use.
• Contributions are welcome. A "Gen AI Pro" plugin is mentioned for production-ready features.
• Community links (Discord/Slack) are not explicitly provided in the README.
• Sponsorship is encouraged to support ongoing development.

Licensing & Compatibility

• The license is not explicitly stated in the provided text. Compatibility for commercial use or closed-source linking is not detailed.

Limitations & Caveats

The plugin is in rapid development and not recommended for production environments. MCP features have known issues, including node connection failures, lack of undo/redo, and improper window docking. Some advanced features like DALL-E, Vision API, Realtime API, and Whisper/TTS are marked as "available in pro" or not yet implemented. Deepseek's reasoning model may require adjusted HTTP timeouts.