gengineEA
Launch App
Beginner

Open the Command Center, configure your AI provider, and send your first message to the Unreal Editor in under 5 minutes.

Quick Start

You have gengine installed and the plugin is loaded. This guide walks you through the Command Center interface, provider configuration, and your first AI-driven editor operation.

Open the Command Center

The Command Center is gengine's main interface inside the Unreal Editor. Open it from the menu bar:

Window > gengine > Command Center

The panel docks like any other Unreal panel — you can drag it to the bottom, float it, or pin it next to the Details panel. It remembers its position between editor sessions.

Tip: Assign a keyboard shortcut to toggle the Command Center via Edit > Editor Preferences > Keyboard Shortcuts > gengine. A shortcut like Ctrl+Shift+G makes it easy to open mid-workflow.

The five tabs

The Command Center has five tabs. Each serves a different purpose:

Chat

The primary interface. Type natural language messages and the AI responds with text and executes tool calls. The Chat tab embeds the Web Chat Panel, which supports topic tabs, persistent memory, asset mentions, and image paste.

Tools

A searchable browser of all 97 MCP operations. Each entry shows the operation name, domain, description, required parameters, and an example call. Use this tab to discover what gengine can do without needing to ask the AI.

Tasks

An async task queue for multi-step operations that run in the background. Each task shows its current status (Queued, Running, Completed, Failed), the sequence of tool calls it made, and the final result. Pro tier users can queue multiple tasks that execute sequentially.

Activity

A real-time feed of every MCP tool call that flows through the system. Each entry shows the tool name, parameters, execution time, and result summary. Useful for understanding what the AI is actually doing behind the scenes.

Diagnostics

Connection health at a glance. Shows the status of the plugin, REST API, MCP bridge, and your configured AI provider. If something is wrong, Diagnostics tells you which component failed and suggests a fix.

Configure your AI provider

Before sending your first message, configure which AI provider gengine should use. Click the gear icon in the top-right corner of the Chat tab (or go to Window > gengine > Settings).

The Provider Settings panel shows a dropdown with supported providers:

  • Claude (Anthropic) — requires the Claude CLI installed and authenticated
  • OpenAI — requires an API key
  • Ollama — runs locally, no API key needed
  • LM Studio — runs locally, requires LM Studio desktop app
  • Custom (OpenAI-compatible) — any provider with an OpenAI-compatible API

Select your provider, fill in the required fields (API key or base URL), and click Test Connection. A green checkmark confirms the provider is reachable and authenticated. See the Provider Setup guide for detailed instructions for each option.

Note: Claude (via the Claude CLI) gives the best results because it natively understands MCP tool calls. GPT-4o and local models also work well for most operations.

Send your first message

With a provider configured, switch to the Chat tab and type your first message. A good starting point:

Spawn a point light at the center of the level (0, 0, 100) and set its intensity to 5000 lux.

Press Enter to send. Watch the Activity tab — you will see two tool calls appear in real time:

  1. unreal_world / spawn_actor — spawns a PointLight actor at (0, 0, 100)
  2. unreal_world / set_property — sets the Intensity property to 5000

The AI replies with a confirmation message and the actor's name in the level. You should see the point light appear in your viewport immediately.

Tip: If the viewport does not update, click inside it and press F to focus the camera on the selected actor.

Use @ mentions for asset references

When you want to refer to a specific asset, type @ followed by the asset name. An autocomplete dropdown appears showing matching assets from your Content Browser.

Create a material instance based on @M_Rock_Base and set the roughness to 0.8

When you select an asset from the autocomplete, gengine inserts its full content path (e.g., /Game/Materials/M_Rock_Base) into the message. The AI receives the exact path and can operate on the asset without guessing.

Asset mentions work for any asset type: Blueprints, Static Meshes, Materials, Textures, Data Tables, Animation Blueprints, and more.

Paste screenshots for visual context

You can paste images directly into the chat with Ctrl+V (or Cmd+V on Mac). This is useful for sharing visual context with the AI — for example, pasting a screenshot of a lighting problem and asking the AI to diagnose it.

1. Take a screenshot (PrintScreen, Snipping Tool, or the viewport capture tool)
2. Click the chat input
3. Press Ctrl+V
4. A preview thumbnail appears in the input box
5. Type your message and send

Images are sent to the AI provider along with your text. Claude and GPT-4o can analyze viewport screenshots to identify lighting issues, actor placement problems, or material appearance discrepancies.

Warning: Image paste requires a multimodal AI provider. Ollama and LM Studio support images only if you have loaded a vision-capable model (e.g., llava, bakllava, or mistral-small with vision).

Best practices

A few habits will make your gengine sessions significantly more productive:

One task per message

The AI executes tool calls sequentially within a message. Breaking complex workflows into separate messages lets you review each step before proceeding, and makes it easier to recover if one step fails.

Be specific about locations and names

Instead of "move that rock", say "move BP_Rock_01 to (500, 200, 0)". Specific actor names and coordinates remove ambiguity and produce correct results the first time.

Use topic tabs to organize sessions

Create a topic per task (e.g., "Lighting Pass", "Level Dressing", "Blueprint Fixes"). Each topic has its own conversation history and persistent memory. Switching topics does not lose your previous context.

Check the Activity tab when something seems off

The Activity tab shows every tool call with its exact parameters and result. If an operation didn't behave as expected, the Activity tab will show exactly what was called and what the editor returned.

Use the Diagnostics tab before asking for help

Most connectivity issues are visible in the Diagnostics tab. Check it first before posting a bug report — it often shows the exact error with a fix suggestion.

← PreviousNext →