Cypress integration

[Lisa-Danielle]

Hey! Any plans to integrate with Cypress?

[miguelg719]

hi @Lisa-Danielle not in the roadmap at the moment but PRs welcome! I'm happy to help you review

[kinthaiofficial]

Stagehand + Cypress integration is an interesting combination — using Stagehand's AI-powered element targeting inside a Cypress test suite.

A few approaches that could work:

Option 1: Stagehand as a CLI fixture generator:

// Use Stagehand to generate reliable CSS selectors/XPaths
// Then use those selectors in Cypress tests
const { Stagehand } = require("@browserbasehq/stagehand");

// Run this once during test setup to generate selectors
async function generateSelectors() {
  const stagehand = new Stagehand();
  await stagehand.init();
  await stagehand.page.goto("https://yourapp.com/login");
  
  // Ask Stagehand to find the right selector
  const loginButton = await stagehand.page.getByAI("the login submit button");
  const selector = await loginButton.evaluate(el => el.getAttribute('data-testid'));
  
  // Write to a fixtures file for Cypress to use
  writeSelectors({ loginButton: `[data-testid="${selector}"]` });
}

Option 2: Stagehand for setup, Cypress for assertions:

// cypress.config.js - run Stagehand setup before tests
const { defineConfig } = require("cypress");

module.exports = defineConfig({
  e2e: {
    setupNodeEvents(on, config) {
      on('before:run', async () => {
        // Use Stagehand to navigate complex setup flows
        // (e.g., through wizard UIs, multi-step onboarding)
        await runStagehandSetup();
      });
    },
  },
});

Option 3: Custom Cypress commands backed by Stagehand:

// cypress/support/commands.js
Cypress.Commands.add('clickByAI', (description) => {
  cy.task('stagehandClick', { description })
    .then(selector => cy.get(selector).click());
});

// Usage in test:
cy.clickByAI('the submit order button in the checkout form');

The value prop for Cypress integration: tests become resilient to UI refactors. Instead of brittle cy.get('[data-cy="submit"]'), you describe intent and Stagehand finds it. The tradeoff is latency (AI call per interaction).

Is the goal faster test authoring, or more resilient existing tests?