This project turns Cursor into an agentic workflow executor to run tasks on your behalf.
The magic is in the .cursor/rules folder. These rules turn Cursor AI into a powerful AI agent that can run scripts and execute tasks on your local machine.
For example, you can ask Cursor to:
- Generate a PDF report from a URL
- Book a flight and hotel
- Resize an image
This project is designed to run on macOS.
- Node.js
- pnpm (TypeScript/Node.js package manager)
- Python
- uv (Python package manager) - 10-100x faster than pip
- Homebrew - macOS package manager
-
Clone this repository
-
Install uv (Python package manager)
-
(Optional) Install media processing tools:
brew install imagemagick ghostscript ffmpeg
-
Install dependencies:
pnpm install
-
Setup your AI provider API keys:
Copy the example environment file:
cp .env.example .env
Recommended MCP servers for Cursor AI. Add them by heading to Cursor Settings > MCP Servers.
{
"mcpServers": {
"browsermcp": {
"command": "npx",
"args": ["@browsermcp/mcp@latest"]
},
"google-maps": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-google-maps"],
"env": {
"GOOGLE_MAPS_API_KEY": "YOUR_GOOGLE_MAPS_API_KEY"
}
}
}
}More servers:
- macos-automator-mcp: A Model Context Protocol server for automating macOS tasks using Automator workflows.
- Peekaboo Tech: A MCP for taking screenshots and more.
- Model Context Protocol Servers List: A comprehensive list of available MCP servers and their documentation.
agent-playground/
├── lib/ # Reusable utilities and shared code
├── scripts/ # Runnable task scripts
├── types/ # Shared TypeScript type definitions
├── data/ # CSV files, JSON data, and other data files
└── package.json # Project configuration
- Open Cursor AI
- Optionally add any relevent files to the
data/directory - Ask Cursor to do its thing
- Profit!