GroundNG: QA for Cursor
Web TestNG MCP
This project provides an AI-powered agent designed to streamline web testing workflows, particularly for developers using AI coding assistants like GitHub Copilot, Cursor, Roo Code, etc. It integrates directly into these assistants via the MCP (Machine Command Protocol), allowing you to automate test recording, execution, and discovery using natural language prompts.
The Problem: Manually testing web applications after generating code with AI assistants is time-consuming and error-prone. Furthermore, AI-driven code changes can inadvertently introduce regressions in previously working features.
The Solution: This tool bridges the gap by enabling your AI coding assistant to:
- Record new test flows: Describe a user journey in natural language, and the agent will interact with the browser (using Playwright) under AI guidance to generate a reproducible test script (JSON format).
- Execute existing tests: Run previously recorded test scripts to perform regression testing, ensuring new code changes haven't broken existing functionality.
- Discover potential test steps: Crawl a website, analyze pages using vision and DOM structure, and ask an LLM to suggest relevant test steps for different pages.
This creates a tighter feedback loop, automating the testing process and allowing the AI assistant (and the developer) to quickly identify and fix issues or regressions.
Server Config
{
"mcpServers": {
"GroundNG-Web-Tester": {
"command": "uv",
"args": [
"--directory",
"path/to/cloned_repo",
"run",
"mcp_server.py"
]
}
}
}