How to Install Funplay Unity MCP via UPM Git URL in Unity 2022.3+

Funplay Unity MCP is a MIT-licensed MCP server for the Unity Editor that exposes 79 tools to AI assistants. You install it through Unity's Package Manager by adding the Git URL https://github.com/FunplayAI/funplay-unity-mcp.git.

What Funplay Unity MCP Is

Funplay MCP for Unity is an open-source Unity Editor MCP server released under the MIT license. It connects AI coding assistants directly to a running Unity project, enabling real-time scene manipulation, script generation, input simulation, and editor automation without leaving your AI workflow.

The server ships with 79 built-in tools organized into two profiles: core with 19 tools and full with all 79. Core covers everyday tasks like scene queries and script creation. Full adds runtime verification, input simulation, performance profiling, and advanced editor automation.

It supports Claude Code, Cursor, VS Code Copilot, Windsurf, and Codex out of the box. The server runs on http://127.0.0.1:8765/ by default.

Prerequisites

• Unity 2022.3 or later
• Git installed on your system and accessible from the command line
• An MCP-compatible AI client such as Claude Code, Cursor, VS Code Copilot, Windsurf, or Codex

Step-by-Step Installation via UPM Git URL

1. Open your Unity project in the Unity Editor.

2. Navigate to Window → Package Manager in the menu bar.

3. Click the + button in the top-left corner of the Package Manager window.

4. Select Add package from git URL... from the dropdown.

5. Paste the following URL into the text field:

https://github.com/FunplayAI/funplay-unity-mcp.git

6. Click Add. Unity resolves the package, downloads it, and imports it into your project. This typically takes a few seconds depending on your network connection.

7. After installation completes, verify the package appears in the Package Manager under the Packages: In Project section.

Starting the MCP Server

Once installed, launch the server from the Unity menu:

1. Go to Funplay → MCP Server in the Unity Editor menu bar.

2. The server starts and listens on http://127.0.0.1:8765/.

3. Confirm the console output shows the server is running and ready for connections.

The server runs as long as the Unity Editor is open. Closing the Editor stops the server.

Configuring Your AI Client

After the server is running, configure your AI assistant to connect to it. The configuration varies by client, but the endpoint is always the same.

For Claude Code, add the MCP server to your configuration file:

{
  "mcpServers": {
    "funplay-unity": {
      "url": "http://127.0.0.1:8765/"
    }
  }
}

For Cursor, navigate to Settings → MCP and add the server URL http://127.0.0.1:8765/.

For VS Code with Copilot, update your MCP settings JSON to include the Funplay server endpoint.

Each client handles tool discovery differently. Most will automatically list the available tools once connected. Switch between the core and full profiles inside Unity to control how many tools the AI sees.

Choosing Between Core and Full Profiles

The core profile exposes 19 tools. Use it for daily development where you want the AI to handle scene queries, basic scripting, and common editor tasks without overwhelming the context window.

The full profile exposes all 79 tools. Switch to this when you need runtime verification, input simulation, performance profiling, or batch editor operations. The larger tool set consumes more context in your AI conversation, so reserve it for tasks that require it.

Switching profiles requires a server restart. Stop the MCP server from the Unity menu, change the profile, then start it again.

Practical Example: Creating a Scene Object via AI

With the MCP server running and your client connected, you can ask your AI assistant to perform Unity tasks directly. For example, using Claude Code after connecting to the Funplay server:

Create a cube at world origin, add a Rigidbody component, and set its mass to 5.

The AI calls the relevant MCP tools to create the GameObject, attach the component, and configure the property — all while you watch it happen in the Unity Editor in real time.

Gotchas and Limitations

Network-only communication. The MCP server communicates over HTTP on localhost. It does not support remote connections out of the box. If you run Unity inside a container or VM, ensure port 8765 is accessible from your AI client.

No hot reload on profile change. Switching between core and full profiles restarts the server. Active AI sessions drop their connection and must reconnect.

Context window cost. The full profile's 79 tools generate a substantial tool definition payload. On smaller context windows, this leaves less room for conversation history and code. Start with the core profile and escalate to full only when necessary.

Unity version lock. The package requires Unity 2022.3 or later. It does not backport to earlier LTS versions. If your project is pinned to Unity 2021 LTS, this tool will not install.

Single-Editor limitation. The server binds to one Unity Editor instance. Running multiple Editor windows on the same machine requires different port configurations, which the current UI does not expose directly.

When Not to Use It

Do not install Funplay Unity MCP if your project operates under a strict no-network policy in the Editor, since the server opens a local HTTP port. Also skip it if you are building for a Unity version older than 2022.3, as the package will fail to resolve.

Next Steps

Clone or star the Funplay Unity MCP repository to receive update notifications, and check out sibling projects under the FunplayAI GitHub organization for MCP servers targeting other engines. For more AI game development guides, visit gamebooom.ai/en/blog/.

meta_title: Install Funplay Unity MCP via UPM Git URL in Unity 2022.3+ meta_description: Learn how to install Funplay Unity MCP with 79 AI tools via UPM Git URL in Unity 2022.3 or later, configure your client, and choose tool profiles. primary_keywords: Funplay Unity MCP,