TornLux/UnrealBridge

Typed control surface for Unreal Engine that lets AI agents introspect assets, author Blueprints/AnimBPs, and edit levels — with reactive events and undoable writes.

What it solves

UnrealBridge solves the problem that AI agents (like code assistants) have no direct, reliable way to inspect and modify Unreal Engine projects. It provides a structured bridge that gives the agent a typed set of operations (via Python functions) to query and edit assets, Blueprint graphs, animations, Niagara effects, and more, with real-time updates and undo support.

How it works

The bridge runs inside the Unreal Editor as a plugin. It starts a local server that the agent's Python client connects to via TCP. The agent sends commands as Python code, which the bridge validates locally (using AST preflight) against a generated manifest of 1388 available functions across 26 libraries, catching mistakes before they reach the editor. Commands execute on the game thread, and results are sent back. The bridge also supports reactive event subscriptions, so the agent can be notified when events occur without polling. A set of bridge libraries covers specific areas like animation, Blueprint editing, Niagara VFX, asset management, and more.

Who it's for

This is for developers and AI-assisted workflows where an LLM-based agent (such as Claude Code, Codex, Gemini CLI, etc.) needs to perform complex tasks in Unreal Engine projects—like creating animation blueprints, setting up Niagara effects, editing Blueprint graphs, or validating gameplay. It's built for interactive editor sessions on Windows, with support for UE 5.3+.

Highlights

  • AST-based hallucination defense: Preflight parses Python code and checks against a manifest, reducing bridge-call failure rate from 24% to 16% in tests.
  • Deep asset-structure introspection and write ops: Libraries for animations, Blueprint graphs, Niagara, and more, with auto-layout and quality tooling.
  • Reactive event subscription: Agent can subscribe to engine events and get callbacks without polling.
  • Native Python execution: 1388 functions across 26 libraries, with all operations wrapped in transactions for undo/redo.
  • Easy integration with Agent Skills: Provides skills for Claude Code and other runtimes following the Agent Skills open standard.

Related

  • Project
    sixtysevenlf/dsh-blender-plugin

    A bridge plugin that enables AI agents to directly control Blender via a TCP channel, allowing for automated scene editing, viewport monitoring, and complex 3D optimization tasks.

  • Project
    ChiR24/Unreal_mcp

    EngineWire is a native Unreal Engine plugin that lets LLM‑based assistants (Claude, Cursor, VS Code, etc.) control the editor through a single MCP tool called `unreal`. It exposes ~400 actions—spawning actors, editing Blueprints, building materials, running Play‑In‑Editor, taking screenshots, and more—via a simple search/describe/execute flow over either a local HTTP server (port 3000) or a Node.js stdio proxy (port 8090). Security is token‑based and localhost‑only by default. Install the plugin, enable the server, add the token to your AI client, and you can issue natural‑language commands that the plugin translates into Unreal‑Editor operations.

  • Project
    unrealcv/unrealcv

    A plugin for Unreal Engine that enables computer vision researchers to build virtual worlds and interact with them using external AI frameworks like PyTorch and TensorFlow.

  • Project
    prajwalshettydev/UnrealGenAISupport

    A plugin for Unreal Engine that simplifies the integration of LLMs and Generative AI, featuring support for multiple AI APIs and the Model Control Protocol for AI-driven editor control.

  • Project
    quodsoler/unreal-engine-skills

    A collection of 31 verified AI agent skills for Unreal Engine 5.8 C++ development, ensuring AI coding agents produce production-quality, up-to-date code.