Unity MCP Server: Build Games With Claude in the Unity Editor (2026)

What Is a Unity MCP Server?

A Unity MCP server connects an AI assistant directly to the Unity Editor. Model Context Protocol is the open standard that lets assistants call external tools, and a Unity MCP exposes the Editor itself as a set of tools: create a GameObject, edit a C# script, set up a scene, import an asset, run the test suite. You describe what you want in natural language, and the assistant performs it inside your live project.

This is a meaningful shift from copy-pasting code out of a chat window. Instead of the AI suggesting a script you then paste into Unity by hand, the AI operates Unity for you: it writes the script, attaches it to the right object, and can run a quick test, all without you leaving the conversation. For solo developers and small studios, that collapses a lot of mechanical Editor work.

Unity MCP servers live alongside other game-engine bridges in the game development MCP servers roundup, and the headline implementation has its own Skiln directory listing.

How the Bridge Actually Works

MCP for Unity is three pieces working together:

• Your MCP client — Claude Desktop, Claude Code, Cursor, Windsurf, Copilot, or a local LLM. This is where you type requests.
• A Python server in the middle — it handles routing, session isolation, and the back-and-forth between the client and Unity.
• A C# Editor plugin inside Unity — it receives routed calls and executes them against the live Editor, then sends results back.

When you ask the assistant to "add a double-jump to the player," the client emits MCP tool calls, the Python server routes them, and the C# plugin edits the script and updates the scene inside Unity. Session isolation in the middle layer is what keeps multiple clients or multiple Editor instances from stepping on each other.

The Unity MCP Servers Worth Knowing

1. MCP for Unity (CoplayDev)

Repo: CoplayDev/unity-mcp · License: MIT · Stars: 5,800-plus · Tools: 25-plus

The de facto standard. MCP for Unity bridges AI assistants with the Unity Editor and gives the LLM tools to manage assets, control scenes, edit scripts, run tests, and automate workflows. It supports multiple clients simultaneously, is automation-ready, and ships under a permissive MIT license. With 5,800-plus stars it is also the most battle-tested option, which matters when you are letting software edit your project files.

Highlights the project calls out: natural-language game development, full Editor control, 25-plus tools, multi-client support, and free MIT licensing. If you install one Unity MCP, install this one. It has a dedicated listing on Skiln so you can track releases.

Best for: almost everyone — solo devs, small studios, and anyone who wants a maintained, popular bridge.

2. mcp-unity (CoderGamester)

Repo: CoderGamester/mcp-unity · Type: Editor plugin

An alternative MCP plugin that connects to the Unity Editor, designed for Cursor, Claude Code, Codex, Windsurf, and other IDEs. It covers similar ground — exposing the Editor to AI clients — and is worth knowing if you want a second option or your workflow centers on a specific IDE. As with any bridge, check recent commit activity and the tool list against your needs before committing your project to it.

Best for: developers who want an alternative, or whose primary tool is Codex or a specific IDE.

Quick Comparison Table

How to Install MCP for Unity

Installation is a Unity package plus a one-click client config:

1. Open Package Manager. In Unity go to Window then Package Manager.
2. Add from git URL. Click the plus button, choose "Add package from git URL," and paste the CoplayDev package URL with the ?path=/MCPForUnity suffix from the repo README.
3. Configure clients. Open Window then MCP for Unity then "Configure All Detected Clients." The plugin finds Claude, Cursor, and other supported clients on your machine and wires them up.
4. Smoke test. In your AI client, ask "list the GameObjects in the open scene." If it answers, the bridge is live.

Because the C# plugin lives in your project, commit it to version control like any other package dependency so teammates get the same setup.

What You Can Build With It

Real tasks the bridge handles well:

• Scaffolding: "Create a first-person controller script and attach it to the Player object."
• Scene work: "Add three patrol-point empties around the enemy and wire them to its waypoint array."
• Refactors: "Split this 400-line GameManager into smaller components."
• Testing: "Run the play-mode tests and summarize any failures."
• Asset management: "Import these textures and create materials for each."

The win is not that any single task is impossible by hand — it is that the assistant chains them, so "prototype a simple platformer level" becomes one request instead of forty clicks.

Limits and Gotchas

Two honest caveats. First, the bridge acts on a live Editor, so a confused instruction can make unwanted changes — version control is not optional, it is the seatbelt. Work on a branch, commit before large automated changes, and review diffs. Second, complex 3D and gameplay design still needs your judgment; the AI is excellent at mechanical Editor operations and boilerplate, less so at deciding whether a mechanic is fun. Treat it like a fast junior developer whose output you still review.

Beyond Unity: Godot and Blender

Unity is not the only engine with an MCP bridge. If you work in Godot, there is a Godot MCP listing that brings the same natural-language control to that engine, and for 3D modeling the Blender MCP lets an assistant drive Blender. Mixing them — Unity for the game, Blender for assets — is a common AI-assisted pipeline. The full set sits in the game development roundup.

Frequently Asked Questions

What is the Unity MCP server?

It is a bridge that connects AI assistants such as Claude, Cursor, Copilot, and Windsurf directly to the Unity Editor through the Model Context Protocol. It gives the assistant tools to manage assets, control scenes, edit scripts, run tests, and automate Editor workflows through natural language. The most popular implementation is CoplayDev/unity-mcp, which is free, MIT-licensed, and has over 5,800 GitHub stars.

Is Unity MCP free?

Yes. CoplayDev's MCP for Unity is open source under the MIT license, free to use and modify. You only need a Unity install and a compatible MCP client. Some hosted or premium AI coding tools that talk to it may have their own pricing, but the bridge itself costs nothing.

How does the Unity MCP bridge work under the hood?

There are three parts: your MCP client (Claude, Cursor, and so on), a Python server in the middle that handles routing and session isolation, and a C# Editor plugin inside Unity. The AI sends an MCP tool call, the Python server routes it, and the C# plugin executes it against the live Editor, then returns the result back up the chain.

Which AI clients work with Unity MCP?

MCP for Unity supports multiple clients including Claude Desktop, Claude Code, Cursor, Windsurf, Copilot, VS Code, and local LLMs. After installing the package you run Window then MCP for Unity then Configure All Detected Clients, and it wires up whichever supported clients it finds on your machine.

How do I install Unity MCP?

In the Unity Editor open Window then Package Manager, click the plus button, choose Add package from git URL, and paste the CoplayDev package URL with the MCPForUnity path. Then open Window then MCP for Unity then Configure All Detected Clients to connect your AI tools. The whole process takes a couple of minutes.

Can Unity MCP write and edit C# scripts?

Yes. The bridge exposes 25-plus tools including ones that create and edit scripts, manage assets, control scenes, and run tests. You can ask the assistant to generate a new player controller, refactor an existing script, or wire up a prefab, and it applies the change directly in the project.

Is it safe to let an AI control my Unity project?

The bridge runs locally and acts on the project you point it at, so the main risk is unwanted edits rather than data exfiltration. Use version control (Git) so every AI change is a reviewable diff you can revert. Work on a branch, commit before large automated changes, and treat the assistant like a fast junior developer whose work you still review.

Last updated: June 30, 2026 · Skiln tracks MCP server release cadence and maintenance weekly.