agent-qa
A self-improving QA agent: natural-language tests for web and mobile that self-heal and remember runs
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Drives the browser and app while running tests
- Runs consume model tokens
Install
Manual install
npx agent-qa init
npx agent-qa install-browsersThen create tests with create-test and run them with run.
This is third-party code. Review the repository files before installing.
What it does
agent-qa is a test harness with memory for teams. Tests are written in natural language for web and mobile: actions and assertions are described in human words, and the agent works from visible roles, labels and screen state. If a single action fails, such as a click or a fill, the agent re-observes the UI and tries a different path in the same run, so tests survive layout drift and flaky steps. Each run builds execution memory that lets the agent adapt to changes and catch regressions before release. It ships a CLI, an MCP server, skills for authoring, debugging and result triage, plus a dashboard and mobile platform support.
Who it is for. For QA engineers and developers who want durable end-to-end tests for web and mobile without brittle selectors.
Good fit when
- You need end-to-end tests written in plain words
- Tests often break due to UI changes
- You want to catch regressions before release automatically
Not a fit when
- You need strictly deterministic tests without an agent
- The project is not about web or mobile interfaces
Example request
Write a test: log in, add an item to the cart and check the counter shows oneLimitations
Tests run through the agent with browsers and mobile drivers, so installed browsers and drivers are needed, and runs consume model tokens. Self-healing lowers brittleness but adds unpredictability compared with hard-coded scenarios. The license is non-standard, so check the terms.
How to disable. Uninstall the agent-qa package and, if used, remove its MCP server from your config.
Security check
- Drives the browser and app while running tests
- Runs consume model tokens
README in short
The README describes agent-qa as an open self-improving QA harness: tests are written in natural language for web and mobile, and the agent works from visible roles and screen state. Execution self-heals on failed steps and builds run memory to adapt to UI drift and catch regressions. It includes a CLI, an MCP server, skills for authoring, debugging and triage, a dashboard and mobile packages. It installs via npm. The license is non-standard.
FAQ
How do tests survive UI changes?
On a failed action the agent re-observes the UI and tries another path in the same run, and run memory helps it adapt.
Web only or mobile too?
Both web and mobile are supported, with separate Android and iOS packages.
Related
A CLI for testing and red teaming LLM applications: compares models, runs automated CI checks and finds vulnerabilities
Playwright CLI
playwright-cli
Microsoft's official Playwright CLI with an agent skill: drive a browser through short commands without heavy MCP schemas
The official MCP server debugger: web UI, automation CLI and terminal UI in one package
SonarSource's official MCP server: quality and security issues, quality gates and code analysis from SonarQube Server and Cloud