At Frontend Nation 2025, Cypress founder Brian Mann took the stage to explore the future of end-to-end testing using natural language. His talk, “Lessons from the AI Battlefield,” cut through the hype and showed what it actually takes to automate browsers with plain English and how Cypress is doing it differently.
Why Natural Language?
Natural language isn’t just a gimmick. It's a powerful tool for:
- Non-developers and QA teams to write tests faster
- Reducing brittleness with “self-healing” selectors
- Increasing portability across teams and systems
- Supporting multilingual prompts for global teams
But Brian didn’t just talk theory. He shared a deep dive into today’s most popular AI browser agents and showed exactly where they fall short for reliable, performant test automation.
What’s the Catch?
Most current tools:
- Aren’t built for testing (they’re built for generic web tasks)
- Hallucinate outputs without inspecting your actual app
- Add serious performance and reliability costs
- Lack support for debugging, assertions, and test lifecycle needs
What’s Cypress Doing About It?
Cypress has been heads-down building something different:
Introducing cy.prompt - a new AI-powered command that lets you describe tests in plain English, right in your test file.
cy.prompt([
`Visit https://aicotravel.com`,
`Type Paris in "where do you want to go"`,
`Click on the first result`,
`Click on the number of day picker`,
`Click on 4 days`,
`Press the Create Itinerary Button`
])It’s fast, it’s resilient, and it adapts as your app evolves without the overhead of traditional AI agents. You can inspect and debug each step of the prompt that was generated.
Want to eject the prompt steps into static test code? You can eject, copy, and save the test code that was generated by cy.prompt to use as you need.
Want Early Access?
We’re rolling out cy.prompt as an experimental feature this summer. Sign up to be among the first to try it.