kao273183/mk-qa-master

mk-qa-master is a universal test runner exposed as a Model Context Protocol, or MCP, server, which means an AI assistant like Claude Desktop or Cursor can use it as a tool to run and understand your test suite directly. Instead of writing tests by hand or manually rerunning them, you point an MCP capable AI client at this server and it can execute tests, look at what failed, and suggest what to fix or write next. It supports a wide range of testing frameworks through a single interface: pytest and Playwright for Python, Jest and Cypress for JavaScript, Go's built in testing tool, Maestro for mobile testing on iOS, Android, and BlueStacks emulators, and as of recent versions, Schemathesis and Newman for testing APIs described by an OpenAPI schema or a Postman collection. Whichever framework is selected, results flow into the same reporting system, so history, flaky test tracking, and coaching advice all work the same way regardless of what kind of test just ran. Beyond running tests, the tool includes an analyzer that can look at a live web page or a mobile app screen and describe its forms, buttons, and navigation elements, which it can then use to generate a runnable test file with real selectors instead of empty placeholders. After a test run, a self improvement coach feature reviews the results across three angles, the quality of the test suite itself, how well the tool is being used, and how effective AI generated tests have been, and produces a prioritized list of what to fix or write next. Installing it is meant to be quick: the recommended path uses a tool called uvx to run the server without a permanent install, needing only a short snippet added to your MCP client's configuration file. A more traditional install into a Python virtual environment is also supported for people contributing to the project. It is released under the MIT license, with documentation in English and Traditional Chinese.