Writing end-to-end tests is still one of the slowest, most error-prone parts of shipping reliable web experiences.
By combining an MCP server with Cursor’s integrated browser control, plus Playwright (an end-to-end test framework for modern web apps), we can now auto-generate and iterate UI tests in minutes (something that would've previously taken days).
Authoring inside the IDE
Cursor - our IDE of choice - now launches and drives a real browser session, observes page state, and writes Playwright specs from concise prompts. This means test creation becomes accessible to everyone, avoiding the bottleneck that usually comes with requiring a specialist.
This enables us to generate “good enough” first passes quickly and frees up more time to spend on high-value edge cases.
Model Discipline and Repeatability
We've noticed that auto-select modes can silently switch models as quotas fluctuate, which can drastically change the quality of output. We get around this by locking down models for repeatable tasks and switching to large-context models only when needed.
Strategic Lens
This is more than a convenience for our developers. It unlocks faster release cadences for our clients and more measurable improvements in customer-facing quality, too.
