Testr

• 

  Landing page

• 

  Dashboard

• 

  Add new test screen

• 

  Test starting up...

• 

  Test in progress...

• 

  Test run review

• 

  Comprehensive analytics with GitHub style activity chart

• 

  API Key screen

• 

  MCP API

• 

  Test run history

• 

  Test Definitions

• 

  Pricing page

🚀 Inspiration

Software development has changed drastically in the past 12 months... but QA hasn't. 😬

We’re shipping faster than ever — yet still forgetting to click through everything before launch. And let’s be real: most small teams (2–5 devs) or solo founders don’t have time, energy, or budget for writing automated test scripts.

That’s where Testr comes in:

• Just describe what to check, and it runs your smoke tests like a human would.
• AI uses browser automation to click around, scroll and input.
• It also live streams its browsing session and in the end gives you a nice summary with a recording.

No pre-recording. No DOM wrangling. Just results.

---

🛠️ What it does

• You write a test like:

  • "test newsletter signup"
  • "check that commenting works"
  • "check that bio opens"

• Testr uses vision-based browser automation to:

  • Click around
  • Scroll intelligently
  • Fill in fields
  • Try to complete the task — like a manual tester would 🧠
  • Live streams its progress straight to UI

• At the end of the test session you get:

  • Clear summary of what succeeded or failed
  • Video recording of the testing session

Bonus: It also comes with a MCP API — so you can automate tests inside your IDE. Just say:

"Please create a test for this view." Testr will generate the prompt based on the actual UI, and run the test.

---

🧱 How we built it

• Mostly built in Bolt.new (which rocks 💥), with some help from Cursor
• Browser automation powered by Puppeteer
• Auth, storage, and tables via Supabase
• Billing with RevenueCat + Stripe
• AI: OpenAI GPT-4o and our own MCP API

---

🧩 Challenges we ran into

Teaching the AI where to click was way harder than expected.

Here’s the journey trying to get it to click “Accept and continue” — red dot = where it actually clicked 🤦

---

🏆 Accomplishments we’re proud of

✅ Browser automation actually works ⚡ It’s fast 🤯 The AI integration + screenshot pipeline came together beautifully

One cool moment: we built a full analytics screen (GitHub-style activity + real data) in Bolt with a single prompt: https://x.com/sgraphics8/status/1935787769416085646

---

📹 Example tests in action

"Test newsletter signup"	"Check that commenting works"	"Check that bio opens"

---

📚 What we learned

AI browser automation isn’t just theory — it’s real, today.

Yes, there’s Playwright MCP, OpenAI Operators, etc... But you can build your own vision-based AI operator from scratch too — using modern tools like Bolt.new and Supabase.

---

⏭️ What’s next for Testr

We're looking for small teams that want to automate smoke tests with AI — and never forget to “click that one button” again 😄

Built With

• bolt.new
• openai
• puppeteer
• revenuecat
• stripe
• supabase