MK QA Master

Universal MCP server that drives your test suite end-to-end: run pytest/Jest/Cypress/Go/Maestro, analyze pages, generate tests, and advise.

Unverified

stdio (local)

No auth

Python

View repo 37 Website

Add to your client

Copy the config for your MCP client and paste it into its config file.

Install / run

uvx mk-qa-master

Paste into ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "mk-qa-master": {
      "command": "uvx",
      "args": [
        "mk-qa-master"
      ],
      "env": {
        "QA_RUNNER": "pytest",
        "QA_PROJECT_ROOT": "/path/to/your-test-project"
      }
    }
  }
}

Requires `uv` (the Python package runner). Install it from https://docs.astral.sh/uv/ if `uvx` is not found.

Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf

Before you start

uv (for the recommended uvx zero-install path) or Python 3 with pip
QA_RUNNER and QA_PROJECT_ROOT environment variables
Per-runner deps: pytest-playwright + 'playwright install chromium' for pytest; a Node project with jest/cypress installed; Go toolchain for go; Maestro CLI + a booted simulator/emulator/device for maestro
API runners: 'pip install mk-qa-master[api]' for Schemathesis; 'npm install -g newman' for Newman
Edge runner: 'pip install mk-qa-master[edge]' + ffmpeg + mediamtx on PATH

About MK QA Master

An MCP server for AI-driven QA: it wires a test framework (pytest+Playwright, Jest, Cypress, Go, Maestro, Schemathesis, or Newman) into Claude/Cursor so the AI can analyze a URL or app screen, generate runnable tests, run them, inspect failures with artifacts, and get a data-driven optimization plan after every run. Exposes a frozen 22-tool surface, stable since v1.0.0.

Tools & capabilities (21)

get_runner_info

Report which test runner is active plus all available runners.

list_tests

Enumerate the tests in the project under test.

run_tests

Run tests with optional filter / headed / browser options (headed and browser are pytest-playwright only).

run_failed

Re-run only the last failures (pytest --lf).

get_test_report

Return a summary: pass / fail / skipped / duration / flaky-in-run.

get_failure_details

Per-failure message plus paths to screenshot, Playwright trace, and video.

generate_test

Generate a test skeleton; with a module from analyze_url/analyze_screen, writes a runnable Playwright .py or Maestro .yaml with concrete selectors.

auto_generate_tests

One-shot: analyze a URL then generate one test per discovered module.

codegen

Launch Playwright codegen (web) or return a hint pointing at 'maestro studio' (mobile).

generate_html_report

Render the latest run as a self-contained HTML report (dark mode).

get_test_history

Return the last N archived run summaries for trend and flake debugging.

analyze_url

Web DOM probe: discovers form/nav/dialog/CTA modules with selectors, candidate test cases, the API endpoints the page hits, and layout-overflow warnings.

analyze_screen

Mobile: runs 'maestro hierarchy' and emits form / cta / tab_bar modules plus noise-filtered candidate test cases.

analyze_stream

Edge AI runner (v1.1.0+): probe RTSP stream geometry and emit candidate detection test cases per detected YOLO label.

init_qa_knowledge

Scaffold the project's bilingual QA knowledge layer (methodology + domain) as qa-knowledge.md.

get_qa_context

Read the QA knowledge layer; optional section filter (e.g. ISTQB, Bug, CAPTCHA, Edge Vision Inference).

get_optimization_plan

Three-layer self-improvement coach: suite quality, MCP usability, and AI generation effectiveness, emitting prioritized actions.

qa_plan

Declare a task plus critical points up front; returns a plan_id threaded into core tools for an auto-verifying plan_verification envelope (universal plan+verify bookend).

inspect_visual_challenge

Detect a reCAPTCHA v2 / hCaptcha image-grid challenge on the active Playwright page and screenshot the tile grid for the AI client. Consent-gated.

solve_visual_challenge

Accept the AI client's tile selection and execute the click chain to solve the CAPTCHA. Requires consent and per-call confirm=true; raw token never returned in evidence.

run_api_security_scan

OWASP API Security Top 10 (2023) rule-based scanner over an OpenAPI 3.x spec: API1 BOLA/IDOR, API2 Broken Auth, API3 Mass Assignment, API5 Function-Level Authz, API8 Misconfig. Consent + authorized-domain gated.

What this server can do

MK QA Master provides tools for these capabilities — tap one to see every MCP server that does the same:

Manage Git repos

When to use it

Analyze a live login or checkout page and auto-generate runnable Playwright tests with concrete selectors
Run a cross-framework test suite (web, mobile via Maestro, or API via Schemathesis/Newman) from a single MCP surface and inspect failures with screenshots and traces
Get a prioritized, data-driven 'what should I fix next' plan after each run, surfacing flaky tests and coverage gaps
Scan an OpenAPI spec for OWASP API Top 10 issues (BOLA, broken auth, mass assignment) on staging environments you own
Drive Maestro flows against iOS Simulator / Android Emulator / BlueStacks via remote ADB without rewriting tests per platform

Security notes

The AI Visual Challenge Solver (reCAPTCHA/hCaptcha) and the OWASP API Security scanner are dangerous capabilities and are disabled by default. They require explicit opt-in env vars: QA_VISUAL_CHALLENGE_CONSENT=true and QA_API_SECURITY_CONSENT=true respectively, plus per-call confirmation. Authorized-domain allowlists (QA_VISUAL_CHALLENGE_AUTHORIZED_DOMAINS, QA_API_SECURITY_AUTHORIZED_DOMAINS) restrict where they operate; hard-stop blacklists refuse known identity providers and camera-vendor hosts. The security scanner runs adversarial test cases and the mass_assignment rule mutates server state — never point it at systems you do not own or are not authorized to test. Secrets in env (auth headers, tokens) are redacted from archived reports by default unless QA_NO_REDACT=1.

MK QA Master FAQ

How do I choose which test framework runs?

Set the QA_RUNNER env var to one of pytest / jest / cypress / go / maestro / schemathesis / newman / edge, and QA_PROJECT_ROOT to the project under test. Each runner has its own prerequisites (e.g. 'playwright install chromium' for pytest-playwright, 'npm install -g newman' for Newman).

Is the CAPTCHA solver enabled by default?

No. Both inspect_visual_challenge and solve_visual_challenge return a consent_required error until you set QA_VISUAL_CHALLENGE_CONSENT=true, and solving requires per-call confirm=true. It also refuses hard-stop domains like known identity providers regardless of consent.

Does it bundle JIRA/Slack/GitHub integrations?

No — it stays a pure test-execution and analysis layer. You compose workflows by running other MCP servers (Atlassian, Slack, GitHub, Sentry, Filesystem) side-by-side in the same client config and letting the AI client orchestrate the chain across servers.

Will accepting updates change the tool surface?

No. mk-qa-master is stable since v1.0.0 with a frozen 22-tool surface enforced by a CI snapshot test. Within v1.x you only get additive changes (new tools, optional args, new fields); breaking changes require a v2.0 bump with a deprecation cycle.

#qa #testing #test-automation #playwright #pytest #jest #cypress #maestro #mobile-testing #api-testing #owasp

Alternatives to MK QA Master

Compare all alternatives →

Playwright MCP Server

Developer Tools

24k

Microsoft's official browser-automation MCP using Playwright's accessibility tree (no vision model).

Featured

Verified

stdio (local)

No auth

TypeScript

12 tools

Updated 21 days agoRepo

Context7 MCP Server

Developer Tools

28k

Up-to-date, version-specific library documentation injected into your coding agent.

Verified

stdio (local)

API key

TypeScript

2 tools

Updated 25 days agoRepo

codebase-memory-mcp

Developer Tools

16k

Code intelligence engine that indexes repos into a persistent knowledge graph for AI coding agents.

Unverified

stdio (local)

No auth

14 tools

Updated 11 hours agoRepo

Compare MK QA Master with:

vs Playwright MCP Server vs Context7 MCP Server vs codebase-memory-mcp vs Serena