Dashboard Overview
Project Create
Test History
Test Video Recording
Sign In

🧠 Visure AI — Autonomous QA for the Next Era of Development

🚀 Inspiration

End-to-end testing is undergoing a transformation. Developers are increasingly combining tools like Cursor AI and Model Context Protocol (MCP) to auto-generate Playwright tests — but something’s missing.

Despite all the advancements, one fact remains: developers still need to write, maintain, and debug tests. And let’s face it — most of us don’t like doing that.

Flaky test failures. Brittle scripts. Mysterious CI breakdowns.

Even with AI in the mix, testing still feels like a chore. Worse, accessibility often falls by the wayside — tagged on late, or skipped entirely.

So I asked:
What if we reinvented end-to-end testing from scratch?
What if developers — or anyone — could test software just by describing what they expect in natural language?

That’s how Visure AI was born. Not another testing tool, but a whole new approach:
Autonomous QA that’s intelligent, accessible, and resilient — powered by language, not code.

💡 What it does

Visure AI is an end-to-end testing assistant that thinks like a QA engineer, not a script runner.

With Visure, you can:

🗣️ Write tests in plain English — no scripts, no boilerplate.
🤖 Auto-generate resilient tests using LLMs + Playwright + MCP.
🧠 Self-heal tests when UI changes, instead of breaking builds.
♿ Automatically scan for accessibility issues and suggest inclusive improvements (Coming Soon).
📹 Record every test session with visual playback and debugging context.
🔁 Run tests async inside CI — GitHub Actions, Bitbucket Pipelines, or webhooks. No agent required.

This isn’t just faster testing.
It’s AI-native quality assurance for the modern development stack.

🛠️ How I built it

Visure AI combines bleeding-edge tools into one cohesive platform:

Playwright — for rock-solid browser automation.
LLMs (Gemini + prompt tuning) — to understand user instructions and generate test logic.
MCP (Model Context Protocol) — customized version of @playwright/mcp@latest with added support for async control and video recording.
CI Integrations — GitHub Actions, Bitbucket Pipelines, and webhook support for agent-free, headless runs.
Netlify — used for hosting the core web interface and test entry flow.

Originally, I used Netlify for hosting and streaming real-time test responses. But I hit platform limitations — especially around long-running serverless functions and request timeouts. So I re-architected the logic to support a "Run in Background" model.

This approach isn’t just more scalable — it’s foundational for fully async CI/CD workflows, where test agents must trigger and run independently of a chat interface or web UI.

Development was powered by Bolt, while I extended the MCP server (via Cursor Background Agents) to support full video capture — ensuring test results are not only seen, but understood.

🛠️ GitHub MCP Fork: https://github.com/lewisvoncken/playwright-mcp

🧱 Challenges I ran into

Like any good testing suite, Visure AI was forged in trial:

🧠 Teaching LLMs to reason through multi-step test logic and DOM state.
🔄 Handling flaky timing conditions without over-engineering wait logic.
📦 Packaging everything into an async CI-ready format that required no interactive interface.
⏱ Designing for performance — minimal latency with maximal reliability.
🎯 Striking the perfect balance between simplicity of input and accuracy of execution.

🏆 Accomplishments that I'm proud of

✅ Built a working system where tests are written in natural language and executed reliably.
📽️ Added full video recording to every test run — debug with confidence, not guesswork.
🧠 Delivered real self-healing test logic to reduce flakiness over time.
🧪 Enabled autonomous QA runs inside CI/CD with zero local setup.
📦 Created a platform that can scale from indie devs to enterprise QA teams.

📚 What we learned

Developers want power, but crave simplicity — AI bridges that gap beautifully.
Test tooling should be inclusive — accessible to QAs, PMs, even non-technical stakeholders.
Generating tests is not enough. What matters is resilience, feedback, and real visibility.
Accessibility isn’t a checkbox — it’s a responsibility. It needs to be first-class.
Async is the future — agents should work where and when the team needs them.

🔮 What’s next for Visure AI

We’re just getting started. Here’s what’s coming soon:

♿ Launching full accessibility analysis with fix suggestions and compliance scoring.
👨‍🔬 Support for test suites, re-runs, and reusable test flows.
🧩 Publishing Visure Agent as an open CI/CD plug-in for all major pipelines.
📱 Expanding into mobile and API-first testing.
💬 Slack/Discord notifications for results, regressions, and a11y reports.
📊 Full dashboard with flaky test tracking, visual diffs, and long-term QA insights.

Visure AI isn’t just a better way to test.
It’s a smarter, more inclusive way to ship software — confidently, and at scale.

🔗 visure.ai