11.5Framework-Native Test Generation: Cypress and Playwright

As of July 2026, natural-language test generation is no longer only a copilot feature -- the test frameworks themselves ship it. If your team is already on Cypress or Playwright, evaluate the framework-native route before adding another tool.

• Cypress v15 cy.prompt() lets you author test steps in natural language directly inside a spec file -- the mainstream, concrete example of NL-to-test. Combined with Cypress Studio and self-healing locators, Cypress has become a genuine AI-authoring player, not just a runner.
• Playwright Test Agents (introduced in v1.56, scaffolded via npx playwright init-agents) provide three cooperating agents: a planner that turns a spec into test scenarios, a generator that writes the Playwright tests, and a healer that repairs them when the UI changes.

The trade-off versus general-purpose copilots: framework-native agents know their framework's idioms deeply (locators, fixtures, auto-waiting) but only generate tests for that framework. A copilot or CLI agent can generate across your whole stack -- unit, API, and E2E. Everything in this chapter about reviewing AI output applies equally to framework-generated tests: the healer that "fixes" a failing test can also silently weaken it, so quality gates still apply.