How QC Pro works

QC Pro runs a real headless browser against your live app, exercises every flow it can find, and produces a report you can hand to Claude Code. Here is what it checks and how the pipeline is structured.

Pipeline

  1. 1

    Crawler

    Discovers your app's route graph, auth walls, and email gates from the URL you paste. No config required.

  2. 2

    Tester

    Plays through discovered flows with a real browser. Captures events, network requests, screenshots, and security observations.

  3. 3

    Analyst

    Scores each of the 12 dimensions and generates findings with evidence, observation, expected behavior, and acceptance criteria.

  4. 4

    Planner

    Groups findings into dependency-ordered phases, each sized to fit a single Claude Code session (~8K tokens).

  5. 5

    Formatter

    Delivers phases via the dashboard, copy-paste, or MCP directly into Claude Code.

The 12 quality dimensions

1. Functional Integrity

What: Do core flows complete without errors?

How: A headless browser signs up with a real disposable email, fills every form, clicks every CTA, and checks that each flow reaches its expected end state without console errors, 4xx/5xx responses, or JavaScript exceptions.

2. Authentication

What: Can users sign up, log in, reset their password, and stay logged in?

How: Tester creates a real account, logs out, logs back in, triggers password reset, and verifies the session persists across reloads and tab changes.

3. Email Notifications

What: Do transactional emails arrive and do their links work?

How: A disposable inbox receives signup/reset/verify mail. The email body is parsed, links are extracted, and each link is followed to confirm it lands on a working page.

4. Responsive / Cross-Browser

What: Does the layout hold across viewports and browser engines?

How: Every primary route renders across 3 mobile, 2 tablet, and 3 desktop viewports, on Chromium and (Pro) WebKit. Overlap and overflow detection runs at each size.

5. Visual Integrity

What: Is the UI visually coherent?

How: Screenshots are analyzed for broken layouts, misaligned elements, clipped text, and images that failed to load. Pixel diff against previous runs catches visual regressions.

6. Performance

What: Does the app load and respond fast enough?

How: Core Web Vitals (LCP, INP, CLS) are measured on every tested route, alongside bundle size, chunk count, and route prefetch overhead.

7. Payment Transactions

What: Do checkout and billing actually work?

How: Tester exercises the checkout flow to the point of payment confirmation (using Stripe test cards only), then verifies the corresponding subscription or order record is created.

8. Third-Party Integrations

What: Do OAuth, analytics, and external APIs actually load?

How: Network interception catches failed or blocked third-party requests. OAuth flows are attempted with test accounts. Embed loads are verified.

9. Security Basics

What: Are there obvious security holes?

How: Scans for XSS, CSRF, missing security headers (CSP, HSTS, X-Frame-Options), exposed API keys in client bundles, insecure cookies, and open redirects.

10. Accessibility

What: Is the app usable with assistive tech?

How: Axe-core runs on every route checking color contrast, keyboard navigation, focus order, ARIA validity, and prefers-reduced-motion support.

11. Regression Detection

What: What broke compared to last time?

How: Each run diffs findings against the previous run for the same URL. New findings are flagged as regressions; fixed ones are marked resolved.

12. Product Intent Alignment

What: Does the app actually do what it claims?

How: Given the stated value proposition, the analyst checks whether the tested flows deliver the promised capability. Gaps are surfaced as high-severity findings.

Claude Code integration

QC Pro exposes an MCP server at https://app.qcpro.ai/api/mcp. Once connected, Claude Code can trigger runs, pull findings, request phased fix instructions, and signal when a phase is ready to retest - all without leaving the editor.

Available MCP tools

  • qcpro/run
  • qcpro/status
  • qcpro/findings
  • qcpro/fix
  • qcpro/retest
  • qcpro/opinion
  • qcpro/history
  • qcpro/intent
  • qcpro/context
Run QC Pro on your appSee example findings