Skip to content
Get packs

Guide · Open source

How to use unity-mcp: install, connect an AI client and prompt Unity

Updated · written with AI from the project's own documentation and checked against our catalogue data

unity-mcp (published as MCP for Unity) is a free, MIT-licensed bridge that lets an AI assistant work inside your Unity Editor. Through the Model Context Protocol, the assistant gets tools to create scenes and GameObjects, edit C# scripts, manage assets, run tests and build. This guide shows you what to install, how to connect an MCP-compatible client, and how to give your first prompt.

MCP for UnityCoplayDev/unity-mcp · ★ 15.6k · MIT

Quick start

  1. 1Check requirements. Confirm your project uses Unity 2021.3 LTS to 6.x and that Python 3.10+ and uv are available.
  2. 2Install the package. In Unity, open Package Manager, choose Add from git URL and paste the MCP for Unity URL, or use the OpenUPM command.
  3. 3Open the configuration window. Go to Window → MCP for Unity in the Unity Editor.
  4. 4Configure clients. Select Configure All Detected Clients so your installed MCP-compatible AI clients are set up.
  5. 5Send a first prompt. Ask your AI client to create a cube at the origin and add a Rigidbody, then check the result.
  6. 6Review changes. Inspect scenes, scripts and assets the AI edits before you keep them.

What you need before installing unity-mcp

The project requires Unity 2021.3 LTS through 6.x and Python 3.10 or newer. Python is used through uv, a Python tool the project links to in its docs.

You also need an MCP-compatible AI client. The README lists Claude Desktop and Claude Code, Cursor, VS Code, Windsurf, Cline and Gemini CLI, among others.

  • Unity 2021.3 LTS to 6.x
  • Python 3.10+ (via uv, https://docs.astral.sh/uv/)
  • An MCP-compatible AI client

Install MCP for Unity in the Unity Package Manager

Open your project in Unity, go to Package Manager and choose Add from git URL. Paste the package URL from the README. The README notes you can pin a release tag, such as #v10.0.0, instead of using #main.

If you prefer OpenUPM, the README gives a second option: add the package with the OpenUPM command.

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

Connect your AI client

After the package is installed, open Window → MCP for Unity in the Unity Editor and select Configure All Detected Clients. This is the setup step the quickstart describes for connecting detected MCP clients.

The project also documents multi-instance routing for running more than one Unity instance, and tool groups for narrowing which tools are enabled. Those details are in the project wiki.

  • Open Window → MCP for Unity
  • Choose Configure All Detected Clients
  • Check that your client is listed as configured

Give your first prompt

Describe the change in plain language in your AI client. The README's example is a request to create a cube at the origin and add a Rigidbody, which the project says appears in seconds.

Start with small, specific requests so you can check each result. AI-generated edits can be wrong and should be reviewed before you rely on them.

  • Create a cube at the origin and add a Rigidbody.
  • Review the created GameObject in the Hierarchy and Inspector.
  • Ask for one change at a time when editing scripts or assets.

Explore the full tool set

The README describes 50 focused MCP tool entrypoints covering scenes, GameObjects, C# scripts, assets, tests, profiling and builds. The full list is in the project's tool catalog on the wiki.

For script validation, the project documents Roslyn Validation. For remote-hosted servers, it documents Remote Server Auth.

  • Tool catalog: https://coplaydev.github.io/unity-mcp/reference/tools/
  • Roslyn Validation guide for script checks
  • Remote Server Auth guide for hosted setups

What people use it for

  • Prototyping a scene by describing objects and their placement in plain language
  • Automating repetitive editor tasks such as asset or component setup
  • Editing C# scripts with an AI assistant and running tests
  • Experimenting with LLM-assisted asset and script changes on a team project

Frequently asked questions

Is unity-mcp free?

Yes. The README states that MCP for Unity is free and MIT-licensed. The license text is in the repository's LICENSE file.

What Unity versions does unity-mcp support?

The README lists Unity 2021.3 LTS through 6.x. Projects on Unity versions outside that range are not covered by the stated requirements.

Do I need Python to use unity-mcp?

Yes. The requirements list Python 3.10 or newer, used via uv. The project describes this as a local Python-based helper that runs alongside the editor.

Which AI clients work with unity-mcp?

The README says it works with any MCP client and names Claude Desktop and Claude Code, Cursor, VS Code, Windsurf, Cline and Gemini CLI, among others.

How do I install unity-mcp?

Use Unity Package Manager's Add from git URL option with the project's package URL, or install through OpenUPM with openupm add com.coplaydev.unity-mcp.

What are alternatives to unity-mcp?

The README lists Godot AI, from the same makers, and mentions other Unity Editor MCP projects. Compare them on supported engine versions and the tools they provide.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

unity-mcp suits Unity developers who already use an MCP-compatible AI client and want to prototype or automate editor tasks from plain-language prompts. It is a poor fit if you need fully predictable, hand-controlled editor changes, use a Unity version outside 2021.3 LTS to 6.x, or don't want a local Python helper running alongside the editor.

Install options & downloads