uac-mcp-server

MCP (Model Context Protocol) server that bridges AI agents to Unity Editor via the Unity API Communicator plugin.

Control Unity Editor from any MCP-compatible AI client — Claude, Cursor, Windsurf, VS Code Copilot, and more.

Features

• 200+ tools auto-discovered from Unity API Communicator endpoints

• 54 curated schemas with typed parameters for core operations (GameObjects, Scenes, Assets, Materials, Prefabs, Build, Console, Camera, Selection)

• Zero config — just install and connect

• Universal — works with any MCP client via stdio transport

Prerequisites

1. Unity API Communicator plugin — install it from unity-api-communicator.com into your Unity project, then open the Unity Editor. The plugin starts an HTTP server on localhost:7777 automatically.

2. Node.js >= 18

This MCP server is a bridge — it requires Unity API Communicator to be installed and Unity Editor to be open. Without it, the MCP server starts and tools are listed, but every tool call returns a connection error.

Installation

Claude Code

claude mcp add --transport stdio unity -- npx -y uac-mcp-server

Claude Desktop

Add to your config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "unity": {
      "command": "npx",
      "args": ["-y", "uac-mcp-server"]
    }
  }
}

Cursor

Go to Settings > MCP Servers > Add new MCP Server, then:

• Name: unity

• Type: command

• Command: npx -y uac-mcp-server

Or add to .cursor/mcp.json in your project:

{
  "mcpServers": {
    "unity": {
      "command": "npx",
      "args": ["-y", "uac-mcp-server"]
    }
  }
}

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "unity": {
      "command": "npx",
      "args": ["-y", "uac-mcp-server"]
    }
  }
}

VS Code (GitHub Copilot)

Add to .vscode/mcp.json in your project:

{
  "servers": {
    "unity": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "uac-mcp-server"]
    }
  }
}

Any MCP Client

The server uses stdio transport. Launch it as a subprocess:

npx -y uac-mcp-server

It reads JSON-RPC 2.0 messages from stdin and writes responses to stdout, following the MCP specification.

Configuration

Pass environment variables via your MCP client config:

{
  "mcpServers": {
    "unity": {
      "command": "npx",
      "args": ["-y", "uac-mcp-server"],
      "env": {
        "UAC_HOST": "localhost",
        "UAC_PORT": "7777"
      }
    }
  }
}

Available Tools

Curated (typed parameters)

Auto-discovered

All additional endpoints from the UAC plugin (Animation, Terrain, Physics, UI, Lights, Navigation, Addressables, Timeline, etc.) are automatically registered with generic parameter schemas. The AI agent uses tool descriptions to determine the correct parameters.

How It Works

AI Client (Claude, Cursor, Copilot, ...)
    │
    │  MCP (stdio, JSON-RPC 2.0)
    ▼
┌──────────────────────────┐
│  uac-mcp-server          │
│  (bridge MCP → HTTP)     │
└───────────┬──────────────┘
            │  HTTP REST (localhost:7777)
            ▼
┌──────────────────────────┐
│  Unity Editor + UAC      │
│  Plugin (200+ endpoints) │
└──────────────────────────┘

On startup, the server calls /api/discover on the UAC plugin to dynamically register all available endpoints as MCP tools. If Unity is not running, it falls back to a built-in static catalog of core tools.

Development

git clone <repo>
cd mcp-server
npm install
npm run build

Test locally:

node dist/index.js

Troubleshooting

Connection refused / tool calls return errors Unity Editor is not open, or UAC is not running. Open your Unity project — the plugin starts automatically.

UAC_PORT mismatch If you changed the UAC port in the plugin settings, pass the matching value:

"env": { "UAC_PORT": "8888" }

Tools not showing up The server calls /api/discover on startup. If Unity starts after the MCP client, restart the MCP server (or the AI client) so it can re-discover tools.

License

MIT