`GrayGooAgent` is a beta-stage, production-oriented Unreal Editor LLM AI assistant built out of our internal studio workflows. It is designed for teams and developers who want AI to participate directly in Unreal Editor work instead of stopping at chat-style suggestions.
Discord Server
It is not a generic AI agent or a general coding agent bolted onto Unreal. `GrayGooAgent` is specialized for Unreal Editor operations, with context assembly, memory behavior, tool discovery, tool parameters, and orchestration tuned specifically for editor-native workflows. That specialization matters because Unreal projects are noisy, highly structured, and full of editor-specific state that generic agents do not handle especially well by default.
It is also not a magic wish machine. Its most reliable value is helping you get up to speed quickly, explore a real project, understand the live editor state, and handle simple but tedious work efficiently. Harder and more open-ended tasks can also work, but results there depend more heavily on model capability, task framing, and available context.
On the workflow side, `GrayGooAgent` can reach directly into a broad set of Unreal Editor surfaces, including Blueprint, Animation Blueprint, UMG, Niagara, asset workflows, scene workflows, and editor automation. In practice, that means it is well suited for tasks such as inspecting and modifying a Blueprint, exploring a project structure, organizing asset work, pushing through small scene changes, gathering information for performance investigation, and moving grounded editor tasks forward instead of only explaining them.
If your work stays inside Unreal Editor, `GrayGooAgent` already stands on its own as a complete primary workflow. It does not require an outside agent to be useful. At the same time, it also works well as an editor-side execution node for larger agent systems. Through its HTTP bridge, outside general agents or coding agents can hand natural-language tasks to the editor while `GrayGooAgent` handles Unreal-specific execution details. This helps isolate editor-heavy context, reduce context pollution, slow context growth, and keep long-running workflows more manageable.
`GrayGooAgent` supports mainstream provider paths, including `OpenAI Plus/Pro(Codex One-click login)`, and it also works with `OpenAI`-compatible endpoints. That means you can pair it with local LLM services as well as other providers that expose the same API shape. The system is implemented in C++ and is designed for responsiveness, iteration speed, and engineering control.
From a safety and control perspective, `GrayGooAgent` does not expose the kind of unrestricted local shell execution you would expect from a traditional terminal agent. Limited command-style inspection, when available, stays behind strict read-only allowlists and path restrictions. Python-backed mutating execution is off by default. We recommend using it with version control so teams can iterate aggressively while keeping change review and rollback in place.
One purchase, long-term use, and C++ source included. `GrayGooAgent` will continue evolving alongside our internal projects, and new workflows and capabilities will continue landing over time.