unity-mcp/README.md

270 lines
8.9 KiB
Markdown

# Unity MCP Package
A Unity package that enables seamless communication between Unity and Large Language Models (LLMs) like Claude Desktop via the **Model Context Protocol (MCP)**. This server acts as a bridge, allowing Unity to send commands to and receive responses from MCP-compliant tools, empowering developers to automate workflows, manipulate assets, and control the Unity Editor programmatically.
Welcome to the initial release of this open-source project! Whether you're looking to integrate LLMs into your Unity workflow or contribute to an exciting new tool, we're thrilled to have you here.
## Overview
The Unity MCP Server provides a bidirectional communication channel between Unity (via C#) and a Python server, enabling:
- **Asset Management**: Create, import, and manipulate Unity assets programmatically.
- **Scene Control**: Manage scenes, objects, and their properties.
- **Material Editing**: Modify materials and their properties.
- **Script Integration**: View, create, and update Unity scripts.
- **Editor Automation**: Control Unity Editor functions like undo, redo, play, and build.
This project is perfect for developers who want to leverage LLMs to enhance their Unity projects or automate repetitive tasks.
## Installation
### Prerequisites
- Unity 2020.3 LTS or newer
- Python 3.7 or newer
- uv package manager
**If you're on Mac, please install uv as**
```bash
brew install uv
```
**On Windows**
```bash
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"
```
and then add to your PATH:
```bash
set Path=%USERPROFILE%\.local\bin;%Path%
```
**On Linux**
```bash
curl -LsSf https://astral.sh/uv/install.sh | sh
```
Otherwise, installation instructions are on their website: [Install uv](https://docs.astral.sh/uv/getting-started/installation/)
**⚠️ Do not proceed before installing UV**
### Unity Package Installation
1. **Add the Unity Package**
- Open Unity Package Manager (`Window > Package Manager`)
- Click the `+` button and select `Add package from git URL`
- Enter: `https://github.com/justinpbarnett/unity-mcp.git`
2. **Set Up Python Environment**
- Navigate to the Python directory in your project:
- If installed as a package: `Library/PackageCache/com.justinpbarnett.unity-mcp/Python`
- If installed locally: `Assets/unity-mcp/Python`
- Install dependencies:
```bash
uv venv
uv pip install -e .
```
### Claude Desktop Integration
1. Open the Unity MCP window (`Window > Unity MCP`)
2. Click the "Configure Claude" button
3. Follow the on-screen instructions to set up the integration
Alternatively, manually configure Claude Desktop:
1. Go to Claude > Settings > Developer > Edit Config
2. Edit `claude_desktop_config.json` to include:
```json
{
"mcpServers": {
"unityMCP": {
"command": "uv",
"args": [
"--directory",
"/path/to/your/unity-mcp/Python",
"run",
"server.py"
]
}
}
}
```
Replace `/path/to/your/unity-mcp/Python` with the actual path to the Unity MCP Python directory.
### Cursor Integration
1. Open the Unity MCP window (`Window > Unity MCP`)
2. Click the "Configure Cursor" button
3. Follow the on-screen instructions to set up the integration
Alternatively, go to Cursor Settings > MCP and paste this as a command:
```bash
uv --directory "/path/to/your/unity-mcp/Python" run server.py
```
Replace `/path/to/your/unity-mcp/Python` with the actual path to the Unity MCP Python directory.
**⚠️ Only run one instance of the MCP server (either on Cursor or Claude Desktop), not both**
4. **Start Claude Desktop or Cursor**
- Launch your preferred tool
- The Unity MCP Server will automatically connect
## Configuration
To connect the MCP Server to tools like Claude Desktop or Cursor:
1. **Open the Unity MCP Window**
In Unity, go to `Window > Unity MCP` to open the editor window.
2. **Configure Your Tools**
- In the Unity MCP window, you'll see buttons to configure **Claude Desktop** or **Cursor**.
- Click the appropriate button and follow the on-screen instructions to set up the integration.
3. **Verify Server Status**
- Check the server status in the Unity MCP window. It will display:
- **Unity Bridge**: Should show "Running" when active.
- **Python Server**: Should show "Connected" (green) when successfully linked.
## Manual Configuration for MCP Clients
If you prefer to manually configure your MCP client (like Claude Desktop or Cursor), you can create the configuration file yourself:
1. **Locate the Configuration Directory**
- **Windows**: `%APPDATA%\Claude\claude_desktop_config.json`
- **macOS**: `~/Library/Application Support/Claude/claude_desktop_config.json`
2. **Create the Configuration File**
Create a JSON file with the following structure:
```json
{
"mcpServers": {
"unityMCP": {
"command": "uv",
"args": [
"--directory",
"/path/to/your/unity-mcp/Python",
"run",
"server.py"
]
}
}
}
```
3. **Find the Correct Python Path**
- If installed as a package: Look in `Library/PackageCache/com.justinpbarnett.unity-mcp/Python`
- If installed locally: Look in `Assets/unity-mcp/Python`
4. **Verify Configuration**
- Ensure the Python path points to the correct directory containing `server.py`
- Make sure the `uv` command is available in your system PATH
- Test the connection using the Unity MCP window
## Usage
Once configured, you can use the MCP Server to interact with LLMs directly from Unity or Python. Here are a couple of examples:
### Creating a Cube in the Scene
```python
# Send a command to create a cube at position (0, 0, 0)
create_primitive(primitive_type="Cube", position=[0, 0, 0])
```
### Changing a Material's Color
```python
# Set a material's color to red (RGBA: 1, 0, 0, 1)
set_material_color(material_name="MyMaterial", color=[1, 0, 0, 1])
```
Explore more commands in the [HOW_TO_ADD_A_TOOL.md](HOW_TO_ADD_A_TOOL.md) file for detailed examples and instructions on extending functionality.
## Features
- **Bidirectional Communication**: Seamlessly send and receive data between Unity and LLMs.
- **Asset Management**: Import assets, instantiate prefabs, and create new prefabs programmatically.
- **Scene Control**: Open, save, and modify scenes, plus create and manipulate game objects.
- **Material Editing**: Apply and modify materials with ease.
- **Script Integration**: Create, view, and update C# scripts within Unity.
- **Editor Automation**: Automate Unity Editor tasks like building projects or entering play mode.
## Contributing
We'd love your help to make the Unity MCP Server even better! Here's how to contribute:
1. **Fork the Repository**
Fork [github.com/justinpbarnett/unity-mcp](https://github.com/justinpbarnett/unity-mcp) to your GitHub account.
2. **Create a Branch**
```bash
git checkout -b feature/your-feature-name
```
3. **Make Changes**
Implement your feature or fix, following the project's coding standards (see [HOW_TO_ADD_A_TOOL.md](HOW_TO_ADD_A_TOOL.md) for guidance).
4. **Commit and Push**
Use clear, descriptive commit messages:
```bash
git commit -m "Add feature: your feature description"
git push origin feature/your-feature-name
```
5. **Submit a Pull Request**
Open a pull request to the `master` branch. Include a description of your changes and any relevant details.
For more details, check out [CONTRIBUTING.md](CONTRIBUTING.md) (to be created).
## License
This project is licensed under the **MIT License**. Feel free to use, modify, and distribute it as you see fit. See the full license [here](https://github.com/justinpbarnett/unity-mcp/blob/master/LICENSE).
## Troubleshooting
Encountering issues? Here are some common fixes:
- **Unity Bridge Not Running**
Ensure the Unity Editor is open and the MCP window is active. Restart Unity if needed.
- **Python Server Not Connected**
- Verify the Python server is running (`python server.py` in the `Python` directory).
- Check `config.json` for correct port settings (default: `unity_port: 6400`, `mcp_port: 6500`).
- Ensure `uv` and dependencies are installed correctly.
- **Configuration Issues with Claude Desktop or Cursor**
Confirm the paths and settings in the configuration dialog match your tool's installation.
For additional help, check the [issue tracker](https://github.com/justinpbarnett/unity-mcp/issues) or file a new issue.
## Contact
Have questions or want to chat about the project? Reach out!
- **X**: [@justinpbarnett](https://x.com/justinpbarnett)
- **GitHub**: [justinpbarnett](https://github.com/justinpbarnett)
- **Discord**: Join our community (link coming soon!).
## Acknowledgments
A huge thanks to everyone who's supported this project's initial release. Special shoutout to Unity Technologies for inspiring tools that push creative boundaries, and to the open-source community for making projects like this possible.
Happy coding, and enjoy integrating LLMs with Unity!