Getting Started

This guide walks you through installing the Unreal MCP Server plugin, verifying it is running, and connecting an AI client.

Prerequisites

• Unreal Engine 5.7
• Windows, macOS, or Linux
• An MCP-compatible AI client (Claude Code, Cursor, VS Code, Windsurf, or similar)

Step 1: Install the Plugin

Copy the UnrealMCPServer/ folder into your project's Plugins/ directory:

YourProject/
  Plugins/
    UnrealMCPServer/
      UnrealMCPServer.uplugin
      Binaries/
      Source/
      Config/
      Bridge/
      ...

Launch the editor. The plugin ships with pre-compiled binaries, so no build step is required — it will load automatically.

For more details and alternative installation methods, see the Installation page.

Step 2: Verify the Server Is Running

The MCP server starts automatically when the editor launches. Look for these messages in the Output Log (Window > Output Log):

LogUnrealMCP: === Unreal MCP Server starting up ===
LogUnrealMCP: Registered 305 tools (preset: Full)
LogUnrealMCP: MCP server ready with 305 tools on port 13579
LogUnrealMCP: Connect your AI client to: http://localhost:13579/mcp
LogUnrealMCP: MCP status bar widget registered
LogUnrealMCP: === Unreal MCP Server startup complete ===

A green status indicator will appear in the editor's bottom toolbar showing MCP: Port 13579 | 305 Tools. If you see a red dot with MCP: Stopped, click it to start the server.

Step 3: Connect Your AI Client

The Unreal MCP Server endpoint is:

http://localhost:13579/mcp

Choose your AI client below for setup instructions.

Claude Code

Run this command in your terminal to register the server:

claude mcp add unreal -t http http://localhost:13579/mcp

Or add it manually to .claude/mcp.json in your project directory:

{
  "mcpServers": {
    "unreal": {
      "type": "url",
      "url": "http://localhost:13579/mcp"
    }
  }
}

Restart Claude Code. Verify the connection with:

claude mcp list

You should see unreal listed with a green checkmark.

Claude Desktop

Claude Desktop uses STDIO transport, so you need the included Python bridge. Edit your Claude Desktop config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "unreal": {
      "command": "python",
      "args": ["C:/path/to/YourProject/Plugins/UnrealMCPServer/Bridge/bridge.py"],
      "env": {
        "UNREAL_MCP_PORT": "13579"
      }
    }
  }
}

Replace the path with the actual location of bridge.py on your system. Restart Claude Desktop after saving.

Prerequisite: Python 3 must be installed. See the STDIO Bridge page for full details.

Cursor

Open Settings > MCP and add a new server, or create .cursor/mcp.json in your project directory:

{
  "mcpServers": {
    "unreal": {
      "url": "http://localhost:13579/mcp"
    }
  }
}

Restart Cursor to pick up the new configuration.

VS Code (Copilot)

Create .vscode/mcp.json in your project directory:

{
  "servers": {
    "unreal": {
      "type": "http",
      "url": "http://localhost:13579/mcp"
    }
  }
}

MCP support in VS Code requires GitHub Copilot with agent mode enabled. Open the Copilot chat panel and switch to Agent mode to use the Unreal MCP tools.

Windsurf

Open Settings > MCP Servers (or Cascade > MCP), and add a new server with the URL:

http://localhost:13579/mcp

Or create ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "unreal": {
      "serverUrl": "http://localhost:13579/mcp"
    }
  }
}

Restart Windsurf to apply.

Antigravity

Open Settings > AI > MCP Servers and add a new HTTP server:

• Name: unreal
• URL: http://localhost:13579/mcp

Or add it to your Antigravity MCP configuration file:

{
  "mcpServers": {
    "unreal": {
      "type": "http",
      "url": "http://localhost:13579/mcp"
    }
  }
}

OpenCode

Add the server to your OpenCode configuration file (~/.config/opencode/config.json):

{
  "mcpServers": {
    "unreal": {
      "type": "http",
      "url": "http://localhost:13579/mcp"
    }
  }
}

Or register it via the CLI:

opencode mcp add unreal http://localhost:13579/mcp

Other MCP Clients

Any MCP-compatible client can connect via Streamable HTTP:

• Endpoint: http://localhost:13579/mcp
• Method: POST (JSON-RPC 2.0 requests)
• Content-Type: application/json

Legacy SSE transport is also supported:

• SSE endpoint: GET http://localhost:13579/sse
• Message endpoint: POST http://localhost:13579/message

For clients that only support STDIO transport, use the included Python bridge. See the STDIO Bridge page.

Step 4: Test the Connection

Once connected, ask your AI client to run a simple command such as:

"List all actors in the current level"

The client will call the list_actors tool and return the results. You can also try:

"Take a screenshot of the viewport"

"Create a PointLight at position 0, 0, 300"

If the connection fails, make sure the Unreal Editor is running and the MCP server is active (green status indicator in the toolbar). See the Installation page for troubleshooting tips.

What to Read Next

• Installation -- Configuration options, INI settings, and troubleshooting
• Tool Categories -- Browse all 305 tools across 42+ categories
• Presets -- Choose the right preset for your workflow
• Resources -- Read-only data sources for project context
• Prompts -- Reusable workflow templates