How does Cypress's architecture differ from Selenium's?

The core difference is where the test code runs.

Selenium: out-of-process control

Your test runs in its own process (Java, Python, C#, JS). Each action is a WebDriver command sent over HTTP to a browser driver, which drives the browser. The test and the app never share memory. That is why Selenium supports many languages and real multi-window scenarios, but every interaction is a network round trip and synchronisation is left to you.

Cypress: in-browser execution

When you run Cypress, it launches a browser and loads the Cypress app, which hosts your application under test in an iframe and runs your spec file in the same browser, in the same JavaScript run loop as the app. Alongside it, a Node.js server process:

• acts as a proxy for all HTTP traffic, which is how cy.intercept can observe and stub requests;
• runs cy.task code and plugins configured in setupNodeEvents (database seeding, reading files);
• manages screenshots, videos and the run.

Because the test sits next to the app, Cypress can:

• retry queries and assertions automatically until they pass or time out, without explicit waits;
• snapshot the DOM at each command so you can time-travel in the runner;
• access window, the app’s objects and localStorage directly;
• control the network and the clock (cy.clock) deterministically.

describe('login', () => {
  it('shows the dashboard', () => {
    cy.intercept('POST', '/api/login').as('login');
    cy.visit('/login');
    cy.get('[data-cy=email]').type('qa@example.com');
    cy.get('[data-cy=password]').type('secret', { log: false });
    cy.get('[data-cy=submit]').click();
    cy.wait('@login').its('response.statusCode').should('eq', 200);
    cy.contains('h1', 'Dashboard').should('be.visible');
  });
});

Trade-offs to mention

• Language: JavaScript or TypeScript only.
• Tabs and windows: Cypress controls one tab. You test links that open new tabs by checking the href or removing target, not by switching windows.
• Origins: a test visiting multiple superdomains needs cy.origin() for the second origin.
• Browsers: Chrome-family, Firefox and Electron are supported; WebKit support is experimental. Selenium covers Safari through safaridriver.
• Events: by default .click() and .type() are simulated in JavaScript, which is very close to real behaviour but not identical to OS-level input. Selenium’s actions go through the browser’s automation interface.

Choosing

For a JavaScript front-end team that wants fast feedback and great debugging on a single-origin app, Cypress is a strong fit. For multi-language organisations, Safari coverage, or flows spanning many windows and domains, Selenium (or Playwright) is often a better fit. I would base the choice on the app’s constraints and team skills, not on popularity.

What interviewers look for

• The in-browser vs out-of-process distinction, stated clearly.
• Knowing the Node process exists and what it does.
• Balanced trade-offs rather than “Cypress is faster” or “Selenium is outdated”.