Muggle Test walks through your web product like a real user — signup, navigation, error states, wrong inputs. No code required. Get your website's test report in under 10 minutes. Test scripts persist — replay them after every deploy to catch regressions, not just first-time bugs. Also: Muggle AI Works (MCP skill suite that adds automated QA to AI coding) and Muggle AI Teams (An agent workflow that we used when developing) Free tier. No credit card.

Hey Product Hunt! We redesigned our signup button. Loved how it looked. Tested it — everything worked. Shipped it immediately. Then users started asking: “Where do I sign up?” The button was right there on the page. It just didn’t look like one to them. That kept happening. Our tests told us things worked. Nobody was testing if things felt right — whether users could find the signup button, whether error messages actually helped, whether the flow made sense. So we built Muggle Test. Paste any URL — including Vercel, Netlify, or Cloudflare preview links — and AI walks through your web product like a real user. Not just the expected flow — it enters wrong inputs, tries confusing navigation, triggers error states on purpose. You get screenshots of every step, and when something fails, the likely causes and recommended next steps. Plus a comprehensive test report with an overall health score and all issues ranked by severity — so you know what to fix first. Test scripts persist. You can replay them after every deploy and catch regressions — not just first-time bugs. Results in under 10 minutes. No code. While building Muggle Test, we also open-sourced two tools we use daily: Muggle AI Works — a lightweight MCP skill suite that adds automated QA to Claude Code, Cursor, Codex, and Windsurf. Scan your web product, import your existing test cases or product documents, auto-generate reusable test scripts. When something fails, you get a full evidence chain — diagnosis logs, screenshots, what went wrong and how to fix it. Not just pass/fail. Scripts persist across runs, unlike Playwright MCP where your agent rewrites tests every session. Uses Muggle AI tokens (free tier included). npm install @muggleai/works Muggle AI Teams — type /muggle-ai-teams and describe what you want built. The workflow triages your requirement, routes each part to the right expert agent automatically, and before development starts, tells you what you didn’t think about — the blind spots outside your expertise. When something breaks, it traces the root cause before attempting a fix. Completely free. npm install @muggleai/teams This is our public test launch. Both tools are MIT licensed. Extended Muggle AI Works trials for PH users who want to go deep. Comment or DM.

Stan and I shipped a signup button two months back. Click event fired, page advanced, test suite green. Early users bounced anyway. Took us a week to figure out it didn't read as clickable to them: no contrast, no affordance, technically perfect.

Aviation calls these "no-fault failures": every system green, the plane still goes down. Most testing answers "did the code do what we wrote." It doesn't ask "did the user do what they tried." That gap is why Muggle Test exists.

Honest: no mobile yet, and the first run costs more than a scripted suite.

When you shipped your last visible-but-broken thing, was it a click target, a focus state, an empty state, or something weirder?

Muggle Test

Paste a URL. See what your users actually experience

Paste a URL. See what your users actually experience