Introduction
AI writes code faster than any team can review it. QualityMax is the independent system that verifies it.
The verification loop
Section titled “The verification loop”The premise is simple, and it is the reason every part of this product exists: you cannot review your own work. An agent that writes a change and then judges its own change is not verification — it is the same reasoning applied twice. Confidence goes up; correctness does not.
QualityMax closes the loop from the outside:
- Discover — crawl the running application, or import the repository, spec, or existing suite. Ground truth comes from the software as it actually behaves, not from a description of it.
- Generate — turn that understanding into executable tests: Playwright, pytest, Go, Rust, k6 load scripts.
- Execute — run them in an isolated sandbox or on your own machine, and capture what happened.
- Judge — gate the pull request on the result, with review from a model that did not write the code.
- Prove — inspect execution artifacts and, for supported verified-fix workflows, a signed verdict tied to the tested revisions.
- Learn — feed outcomes back into project memory, so the next run starts from what the last one established.
Every capability below is one part of that loop. If you read nothing else, read step 5: evidence is what separates verification from another opinion.
Product matrix
Section titled “Product matrix”The platform is broad. This table is the map — what each line is, and what you actually do with it.
| Product line | What it is | What you do with it |
|---|---|---|
| Discovery & Generation | An AI crawler that explores a live app, plus importers for repos, specs, and documents | Point it at a URL and get a test suite grounded in real pages, forms, and flows |
| Test Management | The system of record — Projects, Test Cases, Automation Scripts | Organise coverage, import from existing tools, export without lock-in |
| Execution | A multi-runtime executor — Playwright, pytest, Go, Rust, Jest | Run suites in a managed cloud sandbox or on your own hardware |
| Self-Healing | Selector scoring and DOM-diff repair | Stop rewriting tests every time the UI shifts |
| Code Review & Gates | Four sequential gates on every pull request, plus a multi-persona review ensemble | Block merges that break behaviour, with review that learns your team’s dispositions |
| Security | Multiple static-analysis engines plus LLM review, and live site audits | Catch vulnerabilities in the diff, before they reach production |
| Performance | A k6 stack — distributed runs, extensions, budgets, and converters | Load-test from CI, and migrate off JMeter, Gatling, or Locust |
| AI Validation | Adversarial and probabilistic evaluation, and simulated end-user personas | Test the parts of an AI product that ordinary assertions cannot reach |
| Mobile | Cloud iOS Simulators and Android Emulators, plus mobile audits | Take a mobile branch to a running, testable app |
| Integrations | Connections to source hosts, CI systems, deploy platforms, Slack, and issue trackers | Put verification where the work already happens — MCP, Slack, GitHub, and CI |
| AI & Models | Multi-model routing, your own model keys, and personal coding seats | Control which model runs, and what it costs you |
| Memory & Grounding | Per-project memory and retrieval grounding | Make each run benefit from every previous one |
| Evidence & Trust | Artifacts, traces, and cryptographically signed verdicts | Inspect the evidence and verify signed records against a trusted signer |
| Deployment | Hosted, on-premises, and fully isolated deployments | Run the platform wherever your data is allowed to live |
| Partner / White-label | A partner API with webhooks, teams, and organisations | Embed QualityMax inside your own product |
| Reference | Tool, API, CLI, and format references | Look up the exact call you need |
Three ways in
Section titled “Three ways in”QualityMax meets you at whichever surface you already work in:
- From a coding agent — connect the MCP server and drive the platform in natural language from Claude Code, Codex, or any MCP-capable client. Start with the MCP quickstart.
- From the web app — import a repository, crawl an app, and manage suites in the UI. Start with the web app quickstart.
- From the terminal or CI — use the command-line agent for scripted and headless runs. Start with the CLI and local-agent quickstart.
- From Slack — mention QualityMax in an approved project channel to inspect authorized records and approve exact test actions. Start with the Slack guide.
Start with Core concepts for the handful of ideas the rest of the documentation assumes.