Getting Started¶
Prerequisites¶
- Node.js 18+
- Unreal Engine 5.4–5.7
- An MCP-capable AI client (Claude Code, Claude Desktop, Cursor, etc.)
1. Run Setup¶
The interactive setup will walk you through:
- Project path — point it at your
.uprojectfile or project directory - Tool categories — choose which categories to enable (use arrow keys + spacebar to toggle)
- Plugin deployment — copies the C++ bridge plugin into your project
- UE plugin enablement — enables required plugins (PythonScriptPlugin, Niagara, PCG, etc.) based on your selections
- MCP client config — detects installed clients and writes the config for you
You can also pass the project path directly:
Restart Required
After setup, restart the editor once so the C++ bridge plugin loads. From then on it starts automatically with the editor.
2. Manual Configuration¶
If you prefer to skip the interactive setup, add this to your MCP client config:
.mcp.json in project root:
claude_desktop_config.json:
The server auto-deploys the plugin and enables required UE plugins on first run.
4. Verify the Connection¶
Ask the AI to run:
You should see the server mode, project path, and editor connection status.
Manual Verification¶
In the editor, open Window > Developer Tools > Output Log and filter on LogMCPBridge:
5. Explore¶
Good first commands to try:
level(action="get_outliner") — see what's in the current level
asset(action="list") — browse project assets
reflection(action="reflect_class",
className="StaticMeshActor") — understand any UE class
demo(action="step", stepIndex=1) — run the Neon Shrine demo
Switching Projects¶
To point at a different .uproject at runtime:
This re-deploys the bridge and reconnects.
Next Steps¶
- Tool Reference — Full list of every tool and action
- Architecture — How the pieces fit together
- Neon Shrine Demo — Walk through a guided demo scene