v0.1.0 Beta Live

The Death of
Broken Tests.

Write tests in natural language, run at Bun's native speed, and let AI automatically update Selectors when the UI changes.

View on GitHubRead Documentation
Live Terminal

The Problem

Traditional E2E tests are incredibly brittle. Change a single class or ID, and the entire test suite collapses.

await page.click('#submit-btn-v2');
Error: locator resolved to hidden element

The Solution

Cognitive Test Runner. Instead of hardcoding selectors in code, you just write your intent in plain Markdown. Our Agent parses it and dynamically finds elements by semantics.

# login-tests.md
- Login with "user@mail.com" and password "pass"
Agent automatically translates to Code

Dual Memory Architecture

Our core features are built on a revolutionary architecture that combines lightning-fast reflexes with deep AI intelligence.

Fast Path (Reflex)

Instant execution speed for known UIs.

Technical Detail
SQLite query with <5ms latency.

Slow Path (Intelligence)

Completely eliminates manual test maintenance.

Technical Detail
Self-healing using GPT-4o when UI changes.

Semantic Memory

No need to write exact words, just the right intent.

Technical Detail
Uses ChromaDB to understand similar commands.

Visual Verification

Validates what HTML code cannot express.

Technical Detail
Checks graphics, charts, and layouts using Vision LLM.

Web Dashboard Experience

A seamless visual workflow from natural language requirements to robust, self-healing test execution.

1. Upload & Parse

Upload your test document (.md, .txt) or paste test cases directly. The AI Parser extracts structured test cases automatically.

2. Review & Configure

Review extracted test cases, set your target URL, and configure your preferred AI provider (OpenAI, Gemini, Claude).

3. Execute & Heal

Run the test plan. The Cognitive Runner uses the Fast Path for known UIs and the Slow Path to self-heal when the UI changes.

4. Dashboard & Reports

View real-time execution logs, track passed/failed steps, and manage your test plans from a unified dashboard.

Developer-First CLI

Powerful Command Line Interface

Not a fan of web dashboards? The AI QA Agent comes with a robust, standalone CLI designed for developers and CI/CD integration.

zsh - ai-qa run
âžœai-qa run ./examples/login-tests-v2.md --url http://localhost:4000
â §AI is parsing test cases from document...
✔ AI parsed 8 test case(s) from 1 file(s)
AI QA Agent CLI v0.1.0
URL: http://localhost:4000Tests: 8 test cases from 1 file(s)Model: gpt-4o-miniBrowser: headless
🚀 Starting test run with 8 test cases...
▶TC-001 — Login with valid credentials8.4s
▶TC-002 — Login with invalid password5.2s
✔ 8 passed | 0 errors
📋 HTML Report: ./.ai-qa/reports/report-abc123.html

Global Install

Install once via npm and run `ai-qa` anywhere across your projects without heavy configuration.

$ npm link

AI-Powered Parsing

Point to any markdown or text file. The CLI uses AI to automatically extract and structure test cases.

$ ai-qa parse ./docs/tests.md --url http://localhost

Headless Execution

Run tests in headless mode for CI/CD pipelines, complete with automatic HTML report generation.

$ ai-qa run ./tests.md --url http://localhost

Instant Reruns

Failed a test? Rerun exactly the same plan ID instantly after fixing the issue.

$ ai-qa rerun <run_id>

Local SQLite Storage

All test plans, runs, and memory are stored locally in a `.ai-qa/` folder inside your project workspace.

View reports in .ai-qa/reports/

Technical Excellence

Built for performance, security, and privacy from the ground up.

Bun 1.1+ Native

Leverages the maximum speed of the most modern TypeScript runtime.

Strict Isolation

Memory is isolated by testPlanId, ensuring no data overlap between projects.

Privacy First

Option to use Local Embedding (all-MiniLM-L6-v2) to ensure absolute data privacy.

Meet the Author

The mind behind AI QA Agent. Passionate about building impactful, scalable digital products.

Lê Việt Anh

Lê Việt Anh

@BrianLe188
Da Nang, Vietnam

🚀 Full-Stack Developer | Aspiring Product Founder

My expertise spans robust backend systems and intuitive, high-performance frontend experiences — with a strong drive to turn innovative ideas into production-ready solutions.

GitHubLinkedInPortfolio