Find the bugs your users hit first.

A real browser loads your site and records every crash, hydration mismatch, broken link and layout collision. Each one comes traced to the file behind it.

Sign in to start · $1.49 per report

  • React #418–#425
  • Source-mapped
  • 4 viewports
  • First vs third party
  • CORS preflight
yoursite.com · Runtime QAexample run
  • uncaughtFirst party

    TypeError: Cannot read properties of undefined (reading 'price')

    checkout.js:212

  • rejectionFirst party

    Unhandled rejection: Failed to fetch /api/cart

    cart.js:48

  • console.errorThird party

    Widget config missing, falling back

    widget.vendor.io

  1. 01LoadReal ChromiumJavaScript on, like a visitor’s browser.
  2. 02ListenConsole & networkErrors and failed requests as they fire.
  3. 03Compare4 viewportsSSIM diff against your baseline.
  4. 04FixCode in your PDFAn AI-written fix next to each finding.

What it catches

The bugs a scraper can’t see.

HTML-only crawlers never run your JavaScript. BugViso runs all of it, then watches what breaks.

Hydration

Server and client disagree? You’ll know where.

React #418–#425, Next.js crashes and Vue/Nuxt mismatches. Public source maps trace each one to the component and line.

Console

Real errors, sorted by who owns them.

Uncaught exceptions and promise rejections, split first party vs. third party, grouped across pages.

Visual regression

Four screens. One diff.

SSIM against your approved baseline. Video, carousels, ads and cookie banners are masked.

Layout

Collisions and clipped text.

Overlapping buttons and text cut off by overflow. Ellipsis and line-clamp count as intentional.

Links & network

Every link, 50 at a time.

404s, soft 404s and failed CORS preflights. Broken internal links come with the page you meant.

Signal, not noise

Nine red lines. Three real bugs.

Flip the switch to see what the report keeps, and why the rest goes.

  • Failed requests counted once
  • Known browser noise dropped
  • Vendor errors capped at −3
  • Same error on 20 pages = 1 entry

9 errors · one page load

  • Uncaught TypeError: Cannot read properties of undefined (reading 'price')
  • Failed to load resource: the server responded with a status of 404 (hero@2x.webp)
  • Unhandled promise rejection: TypeError: Failed to fetch
  • ResizeObserver loop completed with undelivered notifications.
  • GET https://www.googletagmanager.com/gtm.js net::ERR_BLOCKED_BY_CLIENT
  • [Report Only] Refused to load the script 'https://cdn.chat.io/w.js'
  • Service worker registration failed: The operation is insecure.
  • Blocked a frame with origin "https://yoursite.com" from accessing a cross-origin frame.
  • widget.vendor.io: config missing, falling back

Scoring

Every point has a reason.

Every penalty is capped, so one noisy script can’t sink your grade.

  • −3max −20Internal broken linkEach one on your own domain.
  • −4max −10Hydration mismatchPer visible mismatch (#418, #423, #425).
  • −12ceilingFirst-party errorsLog scale. Uncaught errors count double.
  • −8over 5% diffVisual changeOnly against a baseline you approved.
  • −2max −6Overlapping controlsPer in-flow pair. Fixed layers ignored.
  • −3ceilingThird-party errorsVendor scripts can’t sink your grade.

Finding → fix

Stack trace in. Patch out.

AI writes each fix for your actual code, and it ships in the PDF right next to the finding.

example findings

Detected

React #418 in OrderTime

src/components/OrderTime.tsx:18

Code
#418
Pages
6
Penalty
−4

Why it matters

The server rendered 09:41 UTC; the browser rendered 14:41 local. React discards the server HTML and repaints the tree.

Hurts LCP & CLS

Fix · in your PDF

tsx
function OrderTime({ at }: { at: string }) {
  const [label, setLabel] = useState<string | null>(null)
  // Format in the visitor's timezone after hydration
  useEffect(() => setLabel(new Date(at).toLocaleTimeString()), [at])
  return <time dateTime={at}>{label ?? at.slice(11, 16)}</time>
}

Server and client markup match. #418 is gone.

Where it lands

Built for the handoff.

  • PDF with fixes

    Every finding, ranked, with AI-written code.

  • Approve a baseline

    One click makes today’s layout the reference.

  • Weekly re-scans

    Catch the regression the deploy after it ships.

  • CSV & JSON

    Raw findings for your tracker or tests.

FAQ

Developer questions, answered.

What causes React error #418?

React throws #418 when the HTML the server sent does not match what the browser renders during hydration. Common causes are formatting dates in the visitor’s timezone, reading window or localStorage during the first render, Math.random() or Date.now() in markup, and invalid HTML nesting. Move browser-only values into useEffect so the first client render matches the server.

Which frameworks does BugViso detect hydration errors in?

React’s minified error codes #418 to #425 (including the Suspense hydration failures #419, #421, #422 and #424), unminified React messages, Next.js “client-side exception” crashes and Vue/Nuxt hydration mismatch warnings. Next.js, Remix and Gatsby are covered because the errors come from React itself.

Can it point to the line in my source code?

Yes, when your source maps are publicly reachable. BugViso downloads the map for the bundle that threw and traces the minified position back to the original file, line and component. Without public maps you still get the bundle URL and line.

How does the visual regression check work?

Each scan screenshots the page at 1440, 1024, 768 and 390 px wide. The first scan becomes the baseline, and later scans are compared to it with SSIM (structural similarity). Video, carousels, ads and cookie banners are masked so they never count as changes. You approve a baseline from your results, and only a change from an approved baseline lowers the score.

Will third-party scripts flood my report with errors?

No. Errors are split into first party and third party by the host that threw them, and third-party errors can cost at most 3 points. Failed resource loads are counted once as network failures, known browser noise such as the ResizeObserver loop message is dropped, and the same error on many pages is grouped into one entry.

Which links does BugViso check?

Every link and image found on the page, checked 50 at a time. Pages the crawl already loaded successfully are not requested again. Pages that return 200 but read as “not found” are flagged as soft 404s, broken internal links come with the closest live URL as a suggested fix, and sites that block automated requests are marked unverified instead of broken.

Ship it knowing what breaks.

One URL. Results in about five minutes.

Sign in to start · $1.49 per report