██████╗ ██╗   ██╗ █████╗ ██╗     ██╗████████╗██╗   ██╗███╗   ███╗ █████╗ ██╗  ██╗
██╔═══██╗██║   ██║██╔══██╗██║     ██║╚══██╔══╝╚██╗ ██╔╝████╗ ████║██╔══██╗╚██╗██╔╝
██║   ██║██║   ██║███████║██║     ██║   ██║    ╚████╔╝ ██╔████╔██║███████║ ╚███╔╝     ╱|、
██║▄▄ ██║██║   ██║██╔══██║██║     ██║   ██║     ╚██╔╝  ██║╚██╔╝██║██╔══██║ ██╔██╗    (˚ˎ 。7
╚██████╔╝╚██████╔╝██║  ██║███████╗██║   ██║      ██║   ██║ ╚═╝ ██║██║  ██║██╔╝ ██╗    |、˜〵
 ╚══▀▀═╝  ╚═════╝ ╚═╝  ╚═╝╚══════╝╚═╝   ╚═╝      ╚═╝   ╚═╝     ╚═╝╚═╝  ╚═╝╚═╝  ╚═╝    じしˍ,)ノ

QualityMax

Independent verification for AI-written code.

Discover user journeys, generate and run tests, review changes, and inspect the evidence behind your next release.

Website · Open QualityMax · Documentation · Standalone tools

This repository is the public product guide and documentation entry point. The hosted platform source is maintained separately. Individual tools below have their own source repositories, licenses, and requirements.

Choose your starting point

I want to…	Start here
Generate coverage and inspect a first completed run	Web app quickstart
Connect my coding agent to hosted QualityMax	MCP quickstart
Use the platform CLI and local execution agent	CLI quickstart
Try local tools without a QualityMax account	Standalone tools guide
Evaluate QualityMax for my team	Team evaluator guide

Hosted workflows require a QualityMax account and use the allowances of your plan. Standalone tools have separate prerequisites; model providers, coding-agent subscriptions, and your own compute can have costs.

From a user journey to reviewable evidence

┌─ FROM JOURNEY TO EVIDENCE ───────────────────────────────┐
│                                                          │
│   01  DISCOVER    Map application journeys               │
│        │                                                 │
│   02  GENERATE    Create cases and scripts               │
│        │                                                 │
│   03  REVIEW      Check intent and assertions            │
│        │                                                 │
│   04  EXECUTE     Run the reviewed tests                 │
│        │                                                 │
│   05  INSPECT     Read results and available artifacts   │
│        │                                                 │
│   06  REFINE      Review repairs if needed; rerun        │
│        │                                                 │
│        └────────> Retain the evidence                    │
│                                                          │
└──────────────────────────────────────────────────────────┘

QualityMax organizes work as Projects → Test Cases → Automation Scripts. A test case records expected behavior; scripts implement that intent for a framework. Inspect the completed execution and its available artifacts before treating a queued job or a generated script as proof. Core concepts

Explore the platform

Capability	What to expect	Documentation
Discovery and generation	Discover application journeys and produce reviewable cases and scripts	Discovery & Generation
PR verification	Review changes, run supported suites, and check deployed previews	Code Review & Gates
Nexus	Specialist reviews with attributed findings and shareable reports	Nexus review
AI application validation	Conversation evaluation, adversarial evaluations, and hallucination checks	AI validation
Agentic Eyes	Simulated browser journeys with observations about usability and friction	Agentic Eyes
Self-healing	Investigate failed tests and review proposed repairs	Self-healing
Slack	Project-aware questions and supported actions through proposals and approval	Slack
Grounded project context	Reuse verified facts and inspect their sources	Memory & Grounding
Test management and portability	Organize cases/scripts, import assets, and exchange QTML	Test management · QTML
Performance	k6 workflows and migration guides for existing performance assets	Performance
Mobile	Responsive readiness audits and managed mobile-flow guidance	Mobile
Evidence	Interpret execution artifacts, signed verdicts, and exposure receipts	Evidence & Trust

Understand the PR gates

Gate	Purpose	Behavior
Alpha	Review the diff and run static security checks	Blocking verdicts can fail the check
Gamma	Run a detected, supported native test suite	Failures can fail the check; skips when there is no applicable suite
Delta	Run QualityMax-generated project scripts	Advisory warnings
Beta	Run browser tests against a deployed preview	Triggered by the successful deployment event; failures can fail the check

Configure your repository's required checks for merge enforcement. Enabled gates, supported frameworks, preview deployment events, and project settings determine what runs. The GitHub App and the explicit GitHub Action have different setup and execution paths. See gate verdicts for the decision rules.

Nexus on-demand specialist review is a separate workflow from automatic PR review. A Slack reply from one active specialist does not imply that a full multi-persona review ran.

Local tools and integrations

┌─ FIND YOUR WORKSPACE ────────────────────────────────────┐
│                                                          │
│   QualityMax                                             │
│   ├── In the browser                                     │
│   │   └── Hosted projects, tests, runs, and reports      │
│   ├── In your terminal                                   │
│   │   ├── qmax-code       Coding and QA agent            │
│   │   └── qmax            Platform CLI / local agent     │
│   ├── In your coding agent                               │
│   │   ├── qmax-mcp        Local browser QA tools         │
│   │   └── Free QA Skills  Reusable QA workflows          │
│   └── In your workflow                                   │
│       ├── GitHub          PR checks and Actions          │
│       └── n8n / Slack     Integrations and context       │
│                                                          │
└──────────────────────────────────────────────────────────┘

Project	Use it for	License / status
qmax-mcp	Four standalone MCP tools: scan, inspect, generate a Playwright reproduction, run a test	MIT
qmax-code	A QA terminal agent with standalone and hosted-connected modes	Source available, FSL-1.1-ALv2
Free QA Skills	QA workflows inside supported coding agents	Apache-2.0
9lives	Local test repair, reruns, and reviewable diffs	MIT; prototype
Test Grader	Static Playwright test-quality grading	Apache-2.0
Supply Chain Scanner	Python dependency supply-chain checks	Apache-2.0
qmax local agent	Platform CLI and local test execution	Apache-2.0
GitHub Action	Explicit CI workflows with results and report links	MIT
n8n integration	Connect project, test, and performance workflows	MIT
qmax-receipt	Shared signed exposure-receipt schema	MIT

Local MCP versus hosted MCP: starting the standalone qmax-mcp package exposes its local tools without a QualityMax account. To use hosted projects and execution, follow the hosted MCP quickstart or the package's documented proxy mode. They are distinct connection paths.

For qmax-code installation, use the maintained source README and latest release. Advanced terminal commands can be experimental; check the source README for current availability.

See concrete examples

• Local MCP walkthrough: inspection, reproduction, and execution artifacts.
• 9lives repair demo: a local repair workflow.
• Test Grader examples: sample grades and actionable findings.
• n8n regression-to-Slack workflow: an integration template with setup requirements.
• CI example repository: inspect the example and its current run history.
• Self-healing sandbox cookbook: a separate example with its own provider and compute requirements.

These examples demonstrate their named workflows; they are not live production certifications or universal performance benchmarks. The example guide explains what to inspect.

Deployment and embedding

The hosted application is the standard evaluation path. Sovereign is a private preview with design partners; supported topology and operational responsibilities are agreed during evaluation. Deployment guide

For embedding workflows and evaluating branding requirements, see Partner / White-label.

Help and contributions

Use documentation for setup and product behavior, issues for public bugs and documentation feedback, and CONTRIBUTING.md for contributions. Report sensitive issues through SECURITY.md.

This repository's license is separate from the licenses of the individual tools linked above.