[!IMPORTANT] Union is deprecated. Use Unity's official MCP server.

Union ♾️ - The Unity MCP Server

A Model Context Protocol server for Unity

Key Features

• 🖼️ Multimodal Vision: Your agent can see what you see. It can view the scene, look through any camera, watch play mode, and inspect asset thumbnails.

• 🔎 Powerful Search: Go beyond the project panel with simultaneous search across the hierarchy and project assets.

• ✔️ Superior Code Analysis: Leverage Unity's own compiler for code analysis that is more accurate than your agent's linter.

• ⏩ Quick Start: Get running in seconds with a single mcp.json configuration file.

• 🛠️ Extensible: Add your own project-specific tools with minimal boilerplate.

• 📅 Always Current: Kept up-to-date with the latest MCP protocol version — currently 2025-06-18 via the Official MCP C# SDK.

Compatibility

Setup

1. Install node.js

2. Configure mcp.json

{
  "mcpServers": {
    "unity": {
      "command": "npx",
      "args": ["-y", "@nurture-tech/unity-mcp-runner", "-unityPath", "<path to unity editor>", "-projectPath", "<path to unity project>"]
    }
  }
}

This will automatically install the is.nurture.mcp package in your unity project. Feel free to commit those changes to source control.

About the Tools

Meet your Unity AI toolbox.

Known Issues

• The Google External Dependency Manager (EDMU) causes Unity to hang forever on startup when launched via Cursor on Windows. This is under investigation.

• The test_active_scene tool sometimes fails with the error message Maximum call stack size exceeded.

• The search tool occasionally fails with the error message Search index is not ready yet. Please try again later.

Adding Project-Specific Tools

Union uses the official C# MCP SDK.

1. Create a static class to hold your tools. Add the [McpServerToolType] annotation to the class.

2. Declare static methods to implement each tool. Add the [McpServerTool] annotation to each method.

3. Reference the Services directory for examples.

4. You will likely need to quit unity and restart your agent in order for it to see the new tools.

Usage Tips

Here are some tips to get the most out of Union:

• 🚀 Launch through your agent: Always launch Unity through your AI agent's MCP integration. Launching Unity from the Hub will prevent the MCP server from connecting.

• 📂 Per-project setup: If your agent supports it, configure the MCP server in your per-project settings. This allows you to seamlessly switch between Unity projects.

• ⚙️ Command-line arguments: You can pass additional arguments to Unity for advanced scenarios like running in -batchmode or -nographics for CI/CD pipelines. Add a -- separator before the Unity-specific arguments:

  {
    "mcpServers": {
      "unity": {
        "command": "npx",
        "args": [
          "-y",
          "@nurture-tech/unity-mcp-runner"
          "-unityPath",
          "<path to unity editor>",
          "-projectPath",
          ".",
          "--",
          "-batchmode",
          "-nographics"
        ]
      }
    }
  }
  

• ⚠️ Important: Do not use the -logFile command-line argument. The MCP server relies on Unity's standard output for communication.

• 🖥️ Split screen for vision: For tools like screenshot, ensure the Unity editor is visible on your screen otherwise it will end up screenshotting the foreground window.