MiraiTest – Generate Web Tests from Prompts using ChatGPT + Playwright
I built a tool that lets you write test cases in plain English, and it generates + runs them automatically using ChatGPT and Playwright.
Just launched version 1.1.4 — includes a Test Generator UI, smarter AI prediction, and faster element scanning.
🤔 Why I built this
As someone who’s worked in testing and automation for years, I’ve always found this contradiction:
- Testers think in natural language
- But automated tests need code
- And most UI tests break way too often from small DOM changes
I wanted a way to bridge that gap — without requiring testers to become coders or maintain brittle scripts.
So I started hacking on something I needed for myself. That became MiraiTest.
⚙️ What is MiraiTest?
MiraiTest is an AI-powered web testing framework.
You write a simple prompt like:
Login to example.com
Click the profile icon
Verify the logout button is visible
And MiraiTest will:
- Use ChatGPT API to generate Playwright test code
- Execute it automatically
- Save logs and stable test cases
- Output results in a clean HTML report
🆕 What’s new in version 1.1.4
- 🧪 Test Generator UI (CLI is still there if you prefer)
- ⚡ Improved scan speed – better performance on large apps
- 🧠 Smarter AI predictions – better handling of vague/short prompts
- 📄 Debug logs and test case storage included by default
🧠 Tech stack
- LLM: OpenAI's ChatGPT API
- Automation: Playwright
- Wrapper: Custom CLI (Node.js) + file manager
- Output: HTML reporter + JSON logs
🎥 Demo (18 min)
💬 What does “Mirai” mean?
I forgot to explain this in my video, but:
“Mirai” (未来) means future in Japanese.
And honestly — I want testing to feel like it belongs there.
🚀 What’s next
- [ ] Open-source the CLI tool
- [ ] Build a lightweight desktop UI (Electron?)
- [ ] Add GitHub Actions integration
- [ ] Fine-tune LLM prompts via few-shot examples
🤝 I'm open to...
- Feedback on the UX / prompt structure
- Suggestions for devtools / QA integration
- Connecting with others building in AI + automation
Thanks for reading 🙏
You can follow the journey here or on LinkedIn.
P.S. If you’re curious about testing with LLMs, I’d love to talk.