Token导航 LogoToken导航TokenDH.com
研究检索external-servicegithub未标认证来源可访问许可证需确认审计提醒

unity-mcpUnity MCP 搜索

Agent Skill

unity-mcp 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

250

周安装

10

GitHub Stars

1

下载量

81
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:unity-mcp(Unity MCP 搜索)
来源仓库:https://github.com/akillness/oh-my-unity3d
仓库路径:skills/unity-mcp
安装命令:
npx skills add https://github.com/akillness/oh-my-unity3d --skill unity-mcp
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/akillness/oh-my-unity3d --skill unity-mcp

简介

unity-mcp 将 Unity Editor 作为安全可控的执行目标,支持场景、脚本与包管理自动化。

  • 适用于多项目环境下路由命令至正确实例,并集成 Roslyn 脚本校验。
  • 可与 unity_docs 和 unity_reflect 协同使用,增强 C# 代码分析能力。
  • 首次连接需配置编辑器网络与权限白名单,防止未授权访问。
  • 建议在非生产分支上先行验证自动化操作稳定性。

SKILL.md

unity-mcp

Use this skill when Unity Editor automation is the execution surface. The point is not just "connect to MCP", but to use the editor as a safe, inspectable target for scripted scene changes, package operations, script validation, tests, and API checks.

When to use this skill

  • Configuring an AI client to talk to a running Unity Editor over MCP
  • Automating scene, prefab, script, or package changes from an agent workflow
  • Routing commands to the correct Unity instance in a multi-project session
  • Validating generated C# with Roslyn-aware script tooling
  • Using unity_docs and unity_reflect before writing or editing scripts
  • Running editor-side verification with run_tests, get_test_job, and read_console

Instructions

Step 1: Install the Unity package

In Unity Package Manager:

https://github.com/CoplayDev/unity-mcp.git?path=/MCPForUnity#main

Beta builds can target the #beta branch instead. OpenUPM is also supported by the upstream project.

Step 2: Pick transport mode

HTTP is the default and easiest mode:

{
  "mcpServers": {
    "unityMCP": {
      "url": "http://localhost:8080/mcp"
    }
  }
}

Stdio is useful for clients or environments that prefer local process transport:

{
  "mcpServers": {
    "unityMCP": {
      "command": "uvx",
      "args": ["--from", "mcpforunityserver", "mcp-for-unity", "--transport", "stdio"]
    }
  }
}

Use HTTP unless you specifically need stdio. The in-editor quick start remains:

  1. Window > MCP for Unity
  2. Start Server
  3. Configure your client
  4. Look for Connected

Step 3: Know the high-value tools

Current upstream tools include:

  • scene and object control: manage_scene, manage_gameobject, manage_components, manage_prefabs
  • script lifecycle: create_script, manage_script, script_apply_edits, validate_script, manage_script_capabilities
  • UI, rendering, and content: manage_ui, manage_material, manage_texture, manage_shader, manage_vfx, manage_animation, manage_camera, manage_graphics, manage_packages
  • search and verification: find_in_file, find_gameobjects, run_tests, get_test_job, read_console, unity_docs, unity_reflect
  • control and batching: batch_execute, manage_tools, set_active_instance, refresh_unity

Step 4: Route to the right Unity instance

When multiple editors are open:

  1. inspect the unity_instances resource
  2. call set_active_instance with the reported Name@hash
  3. run all subsequent tools against that active instance

Do this before writing code or mutating scenes. Otherwise the agent can edit the wrong project.

Step 5: Prefer verification-aware editing

Before generating or modifying a script:

  1. use unity_docs for the relevant API page
  2. use unity_reflect when the live project type system matters
  3. create or edit the script
  4. run validate_script
  5. inspect read_console
  6. run run_tests where applicable

This reduces hallucinated API usage and makes unity-mcp a real verification surface, not just a write surface.

Step 6: Use batching for repetitive editor work

batch_execute is the throughput tool. Use it for repetitive asset edits, prefab updates, or texture import changes instead of looping one tool call at a time.

Step 7: Enable Roslyn validation when script correctness matters

The upstream project supports one-click Roslyn installation from the Unity window and menu. Use that path first. Manual DLL installation is the fallback.

Examples

Example 1: HTTP quick start

omu "Create a simple player controller scene"
  -> unity-mcp over http://localhost:8080/mcp
  -> manage_scene / create_script / validate_script / run_tests

Example 2: Multi-instance routing

1. read unity_instances
2. set_active_instance MyProject@abc123
3. manage_scene create prototype arena
4. read_console

Example 3: API-safe code generation

1. unity_docs on CharacterController
2. unity_reflect on PlayerController assembly symbols
3. create_script PlayerController.cs
4. validate_script
5. read_console

Best practices

  1. Use HTTP first; use stdio only when the client or environment needs it.
  2. Always confirm the active Unity instance before mutating project state.
  3. Use unity_docs and unity_reflect before writing complex runtime code.
  4. Pair validate_script with read_console; one without the other is incomplete.
  5. Prefer batch_execute for repeated operations.
  6. Enable Roslyn when strict API validation matters.
  7. Keep package management (manage_packages) and script generation in the same verification loop.

References

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

36.61%
按下载量换算30

Claude

29.02%
按下载量换算24

Cursor

16.23%
按下载量换算13

Gemini CLI

8.61%
按下载量换算7

安全审计

Gen Agent Trust Hub

可疑

Socket

通过

Snyk

通过

权限和风险

external-service

该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills