← Semua picks

Tool Review Conditional

Playwright vs Cypress vs Puppeteer: E2E Testing in 2026

Playwright vs Cypress vs Puppeteer for E2E testing in 2026: which one a small Indonesian dev team should actually adopt, and the honest trade-offs.

12 September 2026 · 7 menit ·Use case: Picking one E2E testing tool for a small Indonesian dev team's first real test suite, weighing flakiness, hiring pool, and cost.
PlaywrightCypressPuppeteer

Your test suite is not slow because Playwright or Cypress is slow. It is slow because someone sprinkled sleep(2000) everywhere to stop the flaky failures, and now every run wastes minutes it does not need. That habit is why most small teams quietly abandon E2E testing after a month. So before comparing Playwright vs Cypress vs Puppeteer for E2E testing in 2026, the real question is which tool makes flakiness someone else’s problem — because for a team of four in Serpong shipping daily, a flaky suite is worse than no suite at all.

Here is the short version, then the reasoning.

TL;DR

  • Playwright — the default. Auto-waiting removes the flakiness trap, it is free, and it runs Chromium, Firefox, and WebKit with zero config.
  • Cypress — worth it if you debug tests visually every day, or you lean hard on React component testing.
  • Puppeteer — not an E2E framework. Keep it for scraping, PDF generation, and one-off automation.

What actually changed by 2026

The landscape settled. Playwright won new-project adoption in TypeScript teams, and it is not close. Cypress kept a loyal base, mostly frontend-heavy React shops. Puppeteer never competed for the testing crown — its niche drifted fully toward automation.

Two things are different this year. First, the AI codegen story matured: both Playwright’s codegen and the newer assistant-driven test recorders can now scaffold a working spec from a click-through in a couple of minutes. That lowers the barrier for a junior to write a first test, but it does not fix brittle selectors — you still review what the recorder emits. Second, CI cost became something teams here genuinely count. GitHub Actions minutes are finite, Cypress Cloud raised prices, and a Jakarta or Bandung team on a rupiah budget feels a recurring USD 67/month more sharply than a team in San Francisco.

Side by side

DimensionPlaywrightCypressPuppeteer
PriceFree (open source)Free core, Cloud from ~USD 67/moFree (open source)
Browser enginesChromium, Firefox, WebKitChrome, Firefox, Edge (limited)Chromium, Firefox
Auto-wait / flakinessBuilt-in, per-actionBuilt-in retry-abilityManual (you write it)
Parallel by defaultYes, zero configNeeds Cloud or manual setupManual
Local debuggingUI mode + trace viewer (good)Time-travel runner (excellent)DevTools by hand
Component testingExperimentalMatureNone
TypeScriptFirst-class, zero configGood, some setupGood, some setup
Best fitFull E2E pipelineVisual debugging, RSC-light ReactScraping, PDF, automation

Playwright

Playwright has been my default for new projects since 2025, and the reason is unglamorous: auto-waiting. Every action waits for the element to be actionable — visible, stable, enabled — before it fires. You stop writing sleeps. You stop chasing race conditions between your test and your app’s render. For a team writing its first suite, this is the difference between tests they trust and tests they mute in CI two weeks later.

Cross-browser is free and real. Install @playwright/test, run npx playwright install, and your specs run on Chromium, Firefox, and WebKit. WebKit is the part that matters locally — a big slice of Indonesian mobile traffic is iPhone, and Safari-only layout bugs are a recurring tax. Catching one broken checkout on WebKit before it hits a real user pays for the whole setup.

Parallelization is zero-config. Test files run in parallel by default. A 50-test suite that ran 8 minutes sequentially drops to 2-3 minutes across four workers, with no paid dashboard in the loop.

The honest weakness: local debugging is a notch below Cypress. The trace viewer is powerful — you get a full timeline with DOM snapshots and network — but it takes a moment to learn, and a junior dev onboarding to E2E finds Cypress’s runner more intuitive on day one. And treat codegen output as a draft, never a commit. Its selectors bind too tightly to the current DOM, and they rot the first time someone renames a class.

Cypress

Cypress still wins one thing decisively: the loop of writing and debugging a test. The time-travel runner — hover over each command, see the exact DOM at that step — makes a failing test five times faster to diagnose than reading a trace file. If your team debugs tests interactively every day, that is a real, daily advantage, not a demo feature.

Component Testing is the other genuine edge. Mount a React or Vue component in isolation and test its behavior without booting the whole app. For teams with heavy interactive UI — multi-step forms, a rich text editor, drag-and-drop boards — that is worth a lot on its own, and Playwright has no direct answer.

The trade-offs I weigh seriously: browser support and Cloud pricing. No WebKit means no Safari-engine coverage, which is a gap given the iPhone reality above. Core Cypress runs fine in CI, and you can wire up manual parallelization with something like Mochawesome for reports — but the smooth, managed parallel experience pulls you toward Cypress Cloud, and USD 67/month per three users is a line item a bootstrapped team feels every month.

Pick Cypress if your team lives in Component Testing, or debugs visually all day. For a plain E2E pipeline with no Cloud budget, run the numbers again.

Puppeteer

I respect Puppeteer and I do not recommend it for a test suite — which is not a criticism, because it was never built to be one. It is a browser automation library, not a testing framework.

For scraping and PDF generation, it is still excellent. Direct Chrome DevTools Protocol access gives you low-level control that Playwright and Cypress deliberately hide: intercept network at the raw level, drive a headless browser through an anti-bot flow, render a pixel-accurate PDF from a page. That is Puppeteer’s home turf, and it is a good one.

For E2E testing, you write too much plumbing. No built-in assertions, no native test runner, no automatic retry, no reporter. You can bolt Jest onto Puppeteer and build all of it — but you would be rebuilding what Playwright ships on day one. If you already have a working Puppeteer suite, do not migrate for its own sake. Starting fresh in 2026, choose Playwright.

Verdict

Choose Playwright if you are writing your first serious E2E suite, you need cross-browser coverage including WebKit, you have a budget concern and refuse a paid cloud dependency, or you want auto-waiting to solve flakiness for you.

Choose Cypress if your team leans on mature Component Testing, or someone debugs tests visually every single day and the time-travel runner earns its keep — and you have already priced Cypress Cloud against your CI budget.

Choose Puppeteer if the job is headless automation, scraping, or PDF generation. Not for a CI test suite.

For a small Indonesian team, the safe default is Playwright: it removes the flakiness that kills adoption, adds no recurring dollar cost, and covers the Safari bugs your iPhone users would find first. Cypress remains a valid, conditional pick — just adopt it with eyes open on browser gaps and Cloud pricing, not because a thread told you it felt nicer.

Ditulis oleh Asti Larasati

// Pick Tool Review lain


← Semua picks RSS feed