Evidence for AI-generated code

AI writes the code. See what actually passed.

Turn requirements into tests, run them, and inspect the evidence. QualityMax connects your team’s testing and review workflow; 9lives brings the same evidence-first approach to local Playwright and coding agents.

Explore the demo Book a Demo

Book 15 min directly with a founder — no sales pitch.

Real agent logs. A failing test. A verified fix. No signup.

Or take the 3-min QE Maturity Assessment

QualityMax running a native Android test on a managed emulator

QualityMax running a native iOS test on a managed simulator

Draft tests

Generate tests from repositories, specifications, and observed app behavior. Review the intent before running.

Run checks

Execute tests and inspect their assertions, results, and available artifacts before release.

Review repairs

Inspect the proposed diff and its verification before accepting a test repair.

Inspect evidence

Trace what was checked, what passed, and what still needs verification.

QualityMax × 9lives · Interactive demo

Follow the evidence.
From a broken test to a checked fix.

Explore the testing loop one decision at a time. See how the agent, Playwright, and 9lives work together, then inspect recorded agent runs, console logs, and execution receipts.

01A locator breaksRepair the selector; keep the assertion. 02A green test misses a checkInspect what the test actually establishes. 03Behind the agentObserved controls, bounded actions, receipts. 04Real agent runsScroll through the recorded console evidence. 05Put a fix to the testBefore / after confirmation and injected faults.

Recorded fix confirmation · synthetic profile app

Unfixed revisionAssertion failed

Fixed revisionSame test passed

HTTP 500 · HTTP 401 · malformed JSON3 faults caught

await expect(page.getByRole('status', {name: 'Profile'}))
  .toHaveText('Ready for Ada');

The same assertion fails before the app change and passes after it. Fault injection separately checks that these three failures are detected.

Inspect the patch, logs, and receipts →

Open the full demo ↗ Watch the 60-second launch film ↗ No signup · recorded results · light & dark themes

The first chapters are illustrations. The agent and fix chapters replay real executions on a synthetic local app. They show what those checks establish; they do not run your code or establish complete coverage.

For teams · QualityMax platform

Bring testing into your release workflow.

Generate and manage tests from repositories, specifications, and live apps. Run suites, inspect artifacts, and configure pull request review and quality gates in the hosted platform.

Explore the platform → Compare plans →

For developers & agents · 9lives

Keep Playwright. Add attributable evidence.

Run your existing suite locally with bounded execution and receipts. Assess tests, review verified locator repairs, and use MCP tools from your coding agent. No platform account is required.

Get the local runner ↗ Read the source and limits ↗

From change to coverage

Explore the platform workflow.

This interactive illustration connects a sample spec, test script, run trace, and proposed repair. Its figures are examples, not measured execution receipts. For recorded execution evidence, open the 9lives demo above.

checkout / tc-412 · guest-checkout.spec.ts

tests

guest-checkout.spec.ts
guest_checkout.rs
guest_checkout_test.go
test_guest_checkout.py
guest-checkout.test.js
checkout-load.js
.github.yml
flight_test.mavlink.ts
checkout.swift
CheckoutTest.kt

models

checkout.model.qtml
search-restaurants.model.qtml

Write end-to-end tests in plain English. QualityMax executes your intent directly.

Navigate to the Aurora Hoodie product page

Click "Add to cart"

Proceed to "Checkout as guest"

Fill in shipping address and click pay

TC-412 — Guest can complete checkout. Verify a guest user can purchase a single item end-to-end without signing in.

Steps

Navigate to /products/aurora-hoodie
Add “Aurora Hoodie” to cart
Proceed to checkout as guest
Enter shipping address
Select card payment and submit
Assert order confirmation page is shown

Expected: Order #QM-2048 is created, stock decremented, and a confirmation email is queued.

TC-500 — Peak load checkout flow. Verify system stability with 500 concurrent users checking out.

Scenario

Ramp up to 500 VUs over 1m
Hold at 500 VUs for 3m
Ramp down to 0 over 1m

Expected: HTTP failure rate < 1%, p95 response time < 500ms.

1import { test, expect } from '@playwright/test';
        2
        3test('guest completes checkout', async ({ page }) => {
        4  await page.goto('/products/aurora-hoodie');
        5  await page.getByRole('button', { name: 'Add to cart' }).click();
        6  await page.getByRole('link', { name: 'Checkout as guest' }).click();
        7  await page.fill('[data-test="shipping-name"]', 'Test User');
        8  await page.click('[data-test="pay-card"]');
        9  await expect(page.locator('[data-test="confirmation"]'))
        10    .toBeVisible();
        11 });
        12// generated by qualitymax from tc-412 · self-healing selectors enabled

Generated from tc-412 — selectors use stable data-test attributes with role fallbacks.

Live Execution Feed

Passed run #8841 · chromium · 2m ago

Navigate to product page0.4s
Add “Aurora Hoodie” to cart0.6s
Checkout as guest0.3s
Enter shipping address0.5s
Submit card payment0.2s
Assert confirmation visible0.3s

6/6 steps passed duration 2.3s retries 0 artifacts: trace, video, har

Passed run #8842 · k6-cloud · 5m ago

Peak VUs reached500
Total Requests124,502
HTTP Failures (0.00%)0
p(95) Response Time312ms

Thresholds met duration 5m data transfer 450MB artifacts: html-report, metrics

Selector drift detected on tc-412 step 4. The checkout form was refactored — QualityMax found the new locator and proposes a repair.

Endpoint deprecation detected on tc-500. The /checkout endpoint returned 301 Redirect to /v2/checkout — QualityMax auto-resolved the target URL.

Before (Old Selector)

[ ID: #shipping-name ]

After (New Selector)

[ attr: data-test="shipping-name" ]

-await page.fill('#shipping-name', 'Test User');

+await page.fill('[data-test="shipping-name"]', 'Test User');

proposed [data-test="shipping-name"] confidence 0.96 re-ran locally · passed

proposed /v2/checkout confidence 0.99 re-ran locally · passed

Repair applied · tc-412 passing againRepair applied · tc-500 passing again

The manifesto

Build with whatever you want. Test like it matters.

AI made writing code nearly free. It did not make anything else free — not the 2 a.m. incident, not the churned customer, not the PII in the wrong response body.

Your users can't read your code. They read your product. Shipped PRs aren't the metric anymore, and neither is the landing page. When generating code costs nothing, the only thing you are actually charging for is trust.

The fundamentals of testing haven't changed in 20 years. What changed is how much unverified code now moves past them.

What your users actually see

A checkout that double-charges.
A form that eats their data.
An endpoint that returns someone else's records.

They judge it once. They don't come back to re-judge.

It doesn't matter which tools you built it with. It matters a lot which tools you tested it with.

See what it costs when it goes wrong

The thesis

You cannot review your own work.

AI code generation and AI code testing are adversarial systems that should be separate products. If the same agent writes the code and its tests, the tests pass by construction. They verify what the code does, not what it should do. QualityMax is the independent verifier.

Observed selectors Runnable scripts Sandbox execution CI gates Auditable traces

Not a model wrapper

The moat is not calling Claude or GPT. Anyone can do that. The product is the deterministic pipeline around the AI: page evidence, selectors, assertions, execution artifacts, quality thresholds, and failure records.

An agentic harness

QualityMax lets AI propose tests, then forces those tests through deterministic checks before they affect your merge. The agent can move fast; the harness decides what is allowed to pass.

PROUDLY USED BY

Nestfainder.ai

Anoe.app

OpenDeluxe

ProTest Solutions

Skipper Soft

FOUNDER'S QA BACKGROUND

Deutsche Bank

Cisco

CipherHealth

SemaphoreTddium / Solano CD

The Questions Every Buyer Asks

Answers, not marketing

The questions that stop deals — answered directly, no spin.

What is QualityMax?

QualityMax is an AI test automation platform. It helps teams turn a website, repository, or specification into runnable tests, execute those tests, and inspect the results. It also supports test repair and code review in CI.

Read the product documentation

How does QualityMax check a code change?

QualityMax runs tests and reviews code changes through configurable quality gates. Teams can inspect execution results and available artifacts before deciding to merge. A passing test is evidence for the behavior it checks; it does not prove that every part of a change is correct.

Can I see an example without signing up?

Yes. The public code patch verification example compares an original file, a rejected change, and a successful repair using the same unchanged test. It shows recorded results from a fixed synthetic example. It does not run your code in the browser or prove whole-patch correctness.

Explore the original file, changes, and test results

Will my tests break every time the UI changes?

Self-healing rewrites broken selectors automatically. RAG-grounded generation means selectors come from what your app actually renders — not guesses. When the UI drifts, AI fixes the test without waking you up.

− .btn-submit-v2 broken

+ [data-test="submit"] healed

How long until we have real test coverage?

One repo import → 30+ structured test cases → CI gate. Steps, tags, priorities, and runnable Playwright — generated in minutes. Typical time from zero to CI-gated coverage: under 2 hours.

Import repo 0:00

30+ test cases ~5 min

CI gate live < 2 h

What if Claude or GPT goes down mid-sprint?

We route across Claude, GPT, and Gemini in parallel — not as a fallback. On April 15, 2026, Anthropic had a partial outage. Our customers kept shipping because two other models picked up the work.

Claude active

GPT-4o partial outage

Gemini active

Your CI kept running.

We have no dedicated QA engineer.

QualityMax was built for dev-only teams. AI generates, runs, and heals tests. You review cases, not write selectors. Studio includes hosted AI and concierge setup for production quality gates.

git push

→

AI generates

→

AI heals

→

CI gate

No QA engineer required in the loop.

Can we test our internal app behind a firewall?

Yes. The qmax local agent runs on your machine and reaches apps behind VPNs, firewalls, or on localhost. AI decisions happen server-side; the browser stays inside your network.

Your app private network

↔

qmax your machine

→

AI decisions server-side only

Aren't we just locking into another AI vendor?

Never. Bring your own Claude Code, Codex, or API keys on Free and Vibe. Studio can provide hosted AI, while BYO remains optional.

Is it safe to connect our GitHub account to QualityMax?

We run 7,000+ tests on every commit to our own repo — through QualityMax. GitHub OAuth grants read-only access; we never persist your source code. Dedicated security workflows (SAST, dependency scanning, secret detection) run on every push. We dogfood our own security stack because we have to. Enterprise adds private deployment, SOC 2 audit trail, and SSO.

We already use Claude Code. Why can't we just add skill files for testing?

We do not replace Claude Code. We complement it, especially through qmax-code. Skill files write one test. QualityMax runs the loop. Claude Code has no persistent memory of your app's DOM, no CI trigger, no self-healing when a selector breaks at 3am, and no token optimization. Our RAG pipeline sends focused 800-token selector context instead of full codebases — test generation that costs ~$0.04 via direct prompt runs under $0.004 through our routing (functional tests from imported code; AI-crawl flows vary). The orchestration is the product.

Crawl → Generate → Run → Heal ↩

vs skill files: prompt → one test → done

Why pay for QualityMax when we already pay for Mabl, CodeRabbit, BrowserStack, and SonarQube?

You're already paying ~€1,100/month for tools that don't talk to each other. Mabl covers E2E but not security. SonarQube covers code quality but not browser tests. BrowserStack gives you devices but not AI generation. CodeRabbit reviews PRs but can't run a single test. QualityMax does all of it — E2E, API, security, performance, AI code review — from one import, one CI gate, one invoice at €79/month.

Mabl~€600/moE2E only

SonarQube~€150/mocode quality

BrowserStack~€200/modevices only

CodeRabbit~€150/moreview only

Total~€1,100/mo4 tools · 4 logins · 0 shared context

QualityMax Studio

€79/mo

E2E · API · Security · Performance · AI Review

One Platform. Web, Mobile, API, Performance, Security.

From repository import to production-ready test suites — everything your team needs in one place.

Repository Intelligence

Deep Scan Your Codebase

Import any GitHub repository. QualityMax analyzes your source code, discovers API endpoints, maps frontend components, and identifies coverage gaps — automatically.

Auto-discover API endpoints from source code
Coverage analysis with gap detection
AI-powered test generation by category

Repository insights showing deep scan, coverage tabs, and gap analysis

Test Management

AI-Generated Test Cases at Scale

Generate 30+ test cases from a single repo import. Each test case includes structured steps, expected results, tags, and priority — ready for automation.

Functional, Security, E2E, Smoke categories
Initiate AI Crawl or validate test cases
One-click code generation per test case

Test cases list with 30 AI-generated functional test cases

Code Generation

Production-Ready Test Scripts

Generated Playwright and pytest scripts with real assertions, proper authentication flows, and API endpoint validation. Run, copy, or export instantly.

Playwright JS and Python pytest frameworks
One-click execution with video and screenshots
Push to GitHub with CI/CD workflow included

Generated Playwright test code with syntax highlighting

Project Command Center

Everything Connected

Each project brings together repository data, test frameworks, pipeline runs, quality thresholds, and deployment hooks — one unified view for your entire testing lifecycle.

Quality threshold with configurable gates
Pipeline runs triggered by deploy webhooks
Import from TestRail, Jira, docs, and repos

Project detail view with framework info, pipeline runs, and quality settings

More Capabilities

The operational QA layer around AI-coded software: app maps, repo imports, terminal agents, PR review, security, performance, and CI gates in one system.

Intelligent Test Generation

Describe correct behavior in plain language. QualityMax generates comprehensive test suites automatically.

Import from Docs & Specs

Upload PDFs, Word docs, or spreadsheets. AI extracts test cases from specs in seconds.

Automatic Coverage Discovery

Intelligent crawling maps your app and generates test scenarios, including edge cases humans miss.

Living Test Framework

Tests accumulate into a persistent Playwright framework. Helpers and page objects are shared across your suite.

GitHub CI/CD Pipeline

Push to GitHub with auto-generated workflows. Tests run on every PR with Check Runs and comments.

Deploy Webhooks

Connect Vercel, Netlify, or any platform. Every preview deploy triggers automatic smoke tests.

Quality Gates for Developers

Merge blocked until tests pass. Automated quality gates protect your main branch.

Self-Healing Tests

Intelligent adaptation keeps your quality gates working as your product evolves.

Multi-Model Routing

We don't bet your QA on one model's uptime. Claude, GPT, and Gemini routed per task — when one provider degrades, your tests keep running.

Real Browser Verification

Full browser automation with screenshots, video, and real-time visualization of test runs.

Correctness Validation

Automated quality scoring across functional correctness, security, usability, and accessibility.

AI Agent Safety Testing NEW

Test your AI chatbot for prompt injection, jailbreaks, PII leakage, and bias. 39 adversarial attacks with LLM-as-judge scoring. Free safety scan →

Probabilistic Agent Evals NEW

Run the same eval many times and watch the pass rate refuse to hold still — a distribution with confidence intervals, not one number. Live demo →

Founder Q&A

10 Rapid-Fire Questions on Building QualityMax

Founder Ruslan Strazhnyk on what 20 years in QA taught him about where automation breaks — and why zero-config testing on every PR changes the game for Go, Rust, Python, and Playwright teams.

Watch this video on QualityMax

qmax-code v1.22.0

$ qmax-code

 ██████╗ ███╗   ███╗ █████╗ ██╗  ██╗
██╔═══██╗████╗ ████║██╔══██╗╚██╗██╔╝    /\_/\
██║   ██║██╔████╔██║███████║ ╚███╔╝    ( ^.^ )
██║▄▄ ██║██║╚██╔╝██║██╔══██║ ██╔██╗     > ~ <
╚██████╔╝██║ ╚═╝ ██║██║  ██║██╔╝ ██╗
 ╚══▀▀═╝ ╚═╝     ╚═╝╚═╝  ╚═╝╚═╝  ╚═╝   meow.
                  code 1.22.0

qmax > crawl staging.myapp.com and generate tests█

New category

Meet qmax-code

The first AI testing agent built for the terminal. Nothing else does this — not Copilot, not test platforms, not record-and-replay tools.

50QA Tools

0Config

15MB binary

Crawl, generate, execute & self-heal Playwright + k6 in plain English

Multi-model routing: Claude, OpenAI, Ollama local fallback

In-terminal auto-fix from PR findings · OS-keychain key storage

Open source (Apache-2.0) · BYOLLM · works behind firewalls

See how it works →

The Vibe Coding Tax

A founder vibe-coded a fintech app and shipped it live. API keys sat in the frontend. 175 customers were charged $500 each before he could rotate the keys. One security scan would have caught it.

Trust & Safety

We integrated QualityMax directly into our deployment pipeline and immediately had generated Playwright tests for flows we'd never gotten around to writing manually. Ruslan and the team have built something genuinely useful. We're looking forward to growing with them.

Integration Partner

From my point of view QualityMax looks spectacular. It would especially be helpful for teams who have no dedicated QA — for example, when tests are written exclusively by Software Engineers.

Dmitry D.

Fullstack Developer

QualityMax brings in another level of testability and frontend-testing.

Hagen Hübel

Serial Tech Builder & Chief of Agents, nestfainder.ai

QualityMax secures the verification loop end to end: least-privilege repository access, encrypted credentials, sandboxed execution, audit-ready traces, SAST and secrets scanning, and private deployment options for regulated teams.

QualityMax Trust & Safety

Platform security model

Pricing

Choose the plan that fits your team. Start free, scale as you grow.

Free

€0 /month

For solo devs trying QualityMax. Bring your own LLM keys.

7-day hosted-AI trial · 250 credits + 250 sandbox min
5 Projects · your own Claude Code, Codex, OpenCode, or LLM/API keys
60 min / month Cloud Sandbox · 1 at a time
Unlimited local execution with qmax-code
Deterministic SAST — no LLM security analysis
GitHub & GitLab CI/CD — standard deterministic gates

Get Started

Enterprise / Sovereign

Custom

For regulated teams that need customer-hosted, private-cloud, or hybrid deployment.

Dedicated QA Engineering team
Custom AI Model Training on your code
QualityMax Sovereign customer-hosted control plane
Private Cloud / VPC / air-gapped deployment
Contracted projects, credit pool, and sandbox allowances
Enterprise compliance — SSO / SAML / SCIM, RBAC, SOC 2, audit logs
SLA + Dedicated Success Manager
MSA, DPA, custom procurement
Hosted, private-cloud, or hybrid delivery

Explore Sovereign

One verification loop, different boundaries

Run QualityMax as a hosted platform, fully inside your infrastructure, or against embedded firmware and live telemetry.

QualityMax Sovereign

Keep code, prompts, evidence memory, telemetry, and execution inside customer-controlled infrastructure.

Explore customer-hosted QA

Embedded & Robotics

Test real drone, robot, and IoT firmware in reproducible simulators and verify behavior against live telemetry.

Open the Embedded live proof

Browse the engineering artifact on GitHub

Done for you

Don't want to DIY? We'll do it for you.

Senior AI-security audits (from €6k) and fractional QA leadership (from €3k/mo) — delivered on the QualityMax platform, then left running.

Explore Services →

Free Tool

Is Your AI Agent Safe?

Paste your chatbot URL. We'll attack it 39 ways and tell you what's broken. Free, 60 seconds, no signup required.

Scan Your Agent →