Val vs manual testing and Playwright
There are three common ways to test a web app: click through it yourself, write Playwright or Cypress scripts, or let an AI agent drive a browser for you. Here is what each is good at and when to reach for which.
The short comparison
| Approach | Setup | Coverage | Best for |
|---|---|---|---|
| Manual clicking | None | Whatever you remember to check | A quick eyeball, one happy path |
| Hand-written Playwright | Hours per suite, ongoing maintenance | Exactly what you scripted | A stable regression suite you own |
| Val | One MCP config line | Buttons, forms, a11y, Lighthouse, network, mobile, auth | Exploratory QA driven by your AI, no test code |
Is Val better than writing Playwright tests myself?
They solve different problems. Hand-written Playwright is best for a fixed regression suite you maintain. Val is best for exploratory QA driven by your AI agent in plain language, with no test code to write or maintain. Val uses Playwright under the hood, so you get the same real browser without the scripting.
If you are a team with a long-lived product and a CI pipeline, you probably want both: Val for the fast "is this PR safe to merge" pass, and a hand-written suite for the critical paths you never want to regress. If you are a solo builder shipping a vibecoded app, Val alone covers far more than you would script by hand.
How is Val different from Lighthouse?
Lighthouse gives you a performance and accessibility score for one page. Val runs Lighthouse as one of its checks, then adds button and form exercising, multi-page crawling, login, network capture, and a fix loop with your AI agent. Lighthouse measures; Val measures, finds functional bugs, and helps fix them.
What Val does that the others do not
| Capability | Manual | Playwright (DIY) | Val |
|---|---|---|---|
| No test code to write | yes | no | yes |
| Clicks every button automatically | no | only if scripted | yes |
| Full axe-core a11y audit | no | add-on | yes |
| Full Lighthouse score | no | add-on | yes |
| Catches HTTP 200 with error in body | no | only if scripted | yes |
| Driven by your AI in plain language | no | no | yes |
| Fixes the bugs it finds | no | no | via your agent |
Why use Val instead of just clicking through my app myself?
Manual clicking misses the boring paths: every button on every page, mobile and desktop, behind the login, the network response bodies, the console warnings. Val checks all of it in about 30 seconds and never gets tired or skips a step.
Stop asking your friends to test your app.
Val plugs into Cursor, Claude Code, or Codex. $4.99/month, cancel anytime.
Get Val