vostride

vostride / agent-qa

Public

Open-source Agentic QA Harness with Memory. Write tests in natural language.
 Catch regressions before releases ship.

vostride.comagent-qa agentic-harness agentic-memory agents ai appium
30
1
89% credibility
Found May 20, 2026 at 33 stars 2x -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

agent-qa is an open-source AI-powered testing framework that lets you write automated tests in plain English. Instead of learning complex testing code, you simply describe what your app should do—like 'click the login button' or 'verify the page shows my name.' An AI assistant then reads your instructions and automatically interacts with your web or mobile app, clicking buttons, filling forms, and checking results just like a human tester would. The system is designed to be self-healing: when UI elements change or actions fail, the AI automatically tries different approaches instead of giving up. It also builds execution memory over time, learning from previous runs to make future tests faster and more reliable. You can run tests from the command line or through a polished web dashboard, and you can connect your own AI service (OpenAI, Anthropic, Gemini, or compatible providers) to power the assistant.

How It Works

1
📝 You write tests in plain English

Instead of learning complex code, you describe your test steps in simple sentences like 'Click Login' or 'Verify the page shows Welcome'.

2
🤖 Your AI assistant reads and runs them

agent-qa uses an AI brain to understand your instructions and automatically click, type, scroll, and verify things in your app—just like a human tester would.

3
🔄 It fixes itself when things change

If a button moves or a page looks different, your assistant notices the problem and tries a different approach automatically—no test failures just because of a small UI change.

4
🧠 It gets smarter over time

With every test run, your assistant learns what works and what doesn't. It remembers successful paths so future tests run faster and more reliably.

5
Choose how to run your tests
⌨️
Terminal

Type a command and watch your tests run in real-time with clear output

🌐
Dashboard

Open a web page to see test results, screenshots, and manage your entire QA workflow

6
📱 Test web AND mobile apps

The same simple test language works for websites in your browser and native apps on Android or iOS—no need to write different tests for each platform.

✅ Your tests catch problems before release

Every time you make changes to your app, your AI assistant runs through your test cases and tells you if something broke—no more surprises after deployment.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 33 to 30 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is agent-qa?

Agent-qa is an open-source testing framework that lets you write end-to-end tests in plain English instead of wrestling with selectors and assertion libraries. You describe what you want to happen—"click the login button," "verify the page shows 'Welcome'"—and the agent figures out how to do it. It runs tests against web apps via Playwright and mobile apps via Appium, with built-in self-healing that catches UI drift and retries failed actions automatically. The framework is built in TypeScript and supports Bring Your Own LLM, so you can plug in OpenAI, Anthropic, Gemini, or any OpenAI-compatible endpoint.

Why is it gaining traction?

The memory system is the real differentiator. Unlike traditional test runners that forget everything after each run, agent-qa builds execution context over time—learning which paths work, which healed mid-run, and feeding that back into future executions. The action cache cuts down token usage by reusing validated plans across similar test runs. Hooks run in isolated Docker containers, letting you call APIs, seed fixtures, or tear down state without polluting your test environment. For teams already using coding agents, the MCP server and pre-built skills let those agents author, debug, and triage tests directly.

Who should use this?

QA engineers tired of maintaining brittle selector-based tests will get the most value. Mobile teams testing Android apps should take a close look—the Android adapter handles web views, native UI, and hybrid contexts with dynamic context switching. Teams running LLM-powered coding agents can integrate agent-qa as a tool the agent calls to validate its own work. If you're doing accessibility testing, the built-in WCAG checks are a bonus.

Verdict

Agent-qa is a well-architected CLI with a polished developer experience—init, doctor, dashboard, and run commands all feel production-ready. The credibility score of 0.8999999761581421% reflects a young project with 30 stars, so the ecosystem and community support are still forming. The TypeScript codebase is clean and modular, but adoption should be measured: try it on a non-critical smoke suite first, validate the self-healing behavior against your specific UI patterns, and watch token costs if you're running large test suites against expensive models. Worth evaluating now; maturity will come with community contributions.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.