Unity MCP
A Model Context Protocol server that integrates AI assistants like Claude and Cursor directly into the Unity Editor to automate game development workflows. It provides over 40 built-in tools for scene manipulation and project management, while offering a simple C# API for creating custom extensions.
README
Unity MCP - AI Assistant Integration for Unity Editor
Model Context Protocol (MCP) server that enables AI assistants like Claude, Codex, and Cursor to automate Unity Editor tasks and game development workflows.
Features
- Zero telemetry - Completely private Unity automation. No data collection.
- 40+ built-in tools - Create GameObjects, run tests, build projects, manipulate scenes through AI.
- Simple extension API - Add custom AI tools with a single attribute.
Requirements
- Unity 6 or later
- Any MCP client: Claude Code, Claude Desktop, Codex, Cursor, or other AI assistants
Installation
- Open Unity Package Manager (Window → Package Manager)
- Click
+→ "Add package from git URL" - Enter the URL of whichever version you want
Latest Version (Recommended)
To install the latest version:
https://github.com/Bluepuff71/UnityMCP.git?path=/Package
Specific Version (If I messed something up)
To install a specific version, append #version tag:
https://github.com/Bluepuff71/UnityMCP.git?path=/Package#1.0.0
See Releases for available versions.
Setup
Claude Code
claude mcp add unity-mcp --transport http http://localhost:8080/
Claude Desktop
Add this to your Claude Desktop configuration file:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"unity-mcp": {
"command": "npx",
"args": ["-y", "mcp-remote", "http://localhost:8080/"]
}
}
}
Restart Claude Desktop after saving.
Other MCP Clients (Codex, Cursor, etc.)
Unity MCP runs an HTTP server at http://localhost:8080/. Any MCP client with HTTP transport support can connect directly. For stdio-only clients, use the mcp-remote bridge as shown above.
Note: Configurations for clients other than Claude Code have not been tested. Open a PR!
How Unity Editor Automation Works
Unity MCP uses a native C plugin to maintain an HTTP server on a background thread, independent of Unity's C# scripting domain. This architecture ensures the AI assistant connection remains active during script recompilation.
┌─────────────────┐
│ MCP Client │
│ (Claude, etc.) │
└────────┬────────┘
│ HTTP
▼
┌─────────────────────────────────────┐
│ Native Plugin (C) │
│ - HTTP server on background thread │
│ - Survives domain reloads │
└────────┬────────────────────────────┘
│
▼
┌─────────────────────────────────────┐
│ Unity C# (main thread) │
│ - Executes tools when available │
│ - Returns "recompiling" during │
│ domain reload │
└─────────────────────────────────────┘
When Unity recompiles scripts, the C# domain unloads temporarily. During this time, the native plugin continues accepting HTTP requests but returns a "Unity is recompiling" error instead of disconnecting. Once recompilation completes, requests are forwarded to Unity's main thread for execution.
Adding Custom AI Tools
Create a static method and mark it with [MCPTool]:
using UnityMCP.Editor;
using UnityEngine;
public static class MyCustomTools
{
[MCPTool("hello_world", "Says hello to the specified name")]
public static string SayHello(
[MCPParam("name", "Name to greet", required: true)] string name)
{
return $"Hello, {name}!";
}
[MCPTool("create_cube_at", "Creates a cube at specified position")]
public static object CreateCube(
[MCPParam("x", "X coordinate", required: true)] float x,
[MCPParam("y", "Y coordinate", required: true)] float y,
[MCPParam("z", "Z coordinate", required: true)] float z)
{
GameObject cube = GameObject.CreatePrimitive(PrimitiveType.Cube);
cube.transform.position = new Vector3(x, y, z);
cube.name = "Custom Cube";
return new
{
success = true,
message = $"Created cube at ({x}, {y}, {z})",
instanceID = cube.GetInstanceID()
};
}
}
Tools are automatically discovered on domain reload. No registration needed.
Adding Custom Resources
Resources expose read-only data to AI assistants. Use [MCPResource]:
using UnityMCP.Editor;
using UnityEngine;
public static class MyCustomResources
{
[MCPResource("unity://player/stats", "Current player statistics")]
public static object GetPlayerStats()
{
var player = GameObject.Find("Player");
if (player == null)
return new { error = "Player not found" };
return new
{
position = player.transform.position,
health = player.GetComponent<Health>()?.CurrentHealth ?? 0,
isGrounded = player.GetComponent<CharacterController>()?.isGrounded ?? false
};
}
}
Resources use URI patterns (e.g., unity://player/stats) and are read via resources/read.
License
GPLv3 - see LICENSE file for details.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。