AutoQA

Agent-native lightweight web automated testing toolkit.

AutoQA is designed for a simple workflow:

autoqa scan https://example.com --out ./reports/example --full

An AI agent such as Codex, Hermes, Claude, or another coding agent should be able to run one command, collect test evidence, and produce a usable Markdown report.

Chinese README: README.zh-CN.md

What AutoQA Does

AutoQA is not trying to replace Playwright, Lighthouse, ZAP, or k6. It is a lightweight orchestration and evidence layer that makes these capabilities easier for agents to use.

The first target is practical web testing:

• HTTP availability checks
• redirect chain inspection
• robots.txt validation
• sitemap.xml validation
• public asset checks such as favicon.ico and manifest.json
• title and description checks
• common security header checks
• browser screenshots
• console error collection
• network failure collection
• authenticated page testing with test accounts
• Markdown and JSON reports
• evidence directory for screenshots, headers, HTML previews, logs, and traces

Design Goals

• One command should be enough for a useful default scan.
• Configuration should be optional, not required.
• Reports should be readable by humans and structured enough for agents.
• Evidence should be saved and traceable.
• Authentication should support test accounts without leaking secrets.
• The default mode should be safe and read-only.
• The implementation should stay lightweight and avoid becoming a heavy test management platform.

Planned Architecture

AutoQA is planned as a small multi-language toolkit:

Layer	Stack	Responsibility
Core	Go	CLI, HTTP scanner, concurrent checks, evidence store, result aggregation, MCP server
Browser	TypeScript + Playwright	browser automation, screenshots, traces, console logs, network logs, E2E execution
Agent	Python	result analysis, test planning, report writing, failure explanation, LLM integration

Users and agents should only need to interact with one command:

autoqa

The internal Go, TypeScript, and Python components should be hidden behind that command.

Planned Installation

The exact installation method is still being designed. The intended experience is:

curl -fsSL https://autoqa.dev/install.sh | sh

Or with Docker:

docker run --rm \
  -v "$PWD/reports:/reports" \
  ghcr.io/nesoriel/autoqa:latest \
  scan https://example.com --out /reports/example --full

During early development, the project may require running the Go CLI, TypeScript browser runner, and Python agent components from source.

Planned Usage

Run a full default scan:

autoqa scan https://example.com --out ./reports/example --full

Run a faster smoke scan:

autoqa scan https://example.com --out ./reports/example --profile smoke

Run an SEO-focused scan:

autoqa scan https://example.com --out ./reports/example --profile seo

Use a config file:

autoqa scan --config autoqa.yaml

Expected output:

reports/example/
  report.md
  result.json
  summary.json
  autoqa.log
  evidence/
    screenshots/
    http/
    console/
    network/
    pages/

Test Account Support

AutoQA must support authenticated testing through test accounts. Passwords should not be passed directly on the command line.

Recommended pattern:

export AUTOQA_TEST_PASSWORD='your-test-password'

autoqa scan https://example.com \
  --out ./reports/example \
  --full \
  --login-url /login \
  --username test@example.com \
  --password-env AUTOQA_TEST_PASSWORD

AutoQA should also support Playwright storage state:

autoqa auth login https://example.com \
  --login-url /login \
  --username test@example.com \
  --password-env AUTOQA_TEST_PASSWORD \
  --save-state .autoqa/auth/state.json

autoqa scan https://example.com \
  --out ./reports/example \
  --full \
  --auth-state .autoqa/auth/state.json

Sensitive values such as passwords, cookies, tokens, Authorization, Cookie, and Set-Cookie must be redacted from reports and evidence by default.

Planned Commands

autoqa scan <url>
autoqa scan <url> --out <dir>
autoqa scan <url> --full
autoqa scan <url> --profile smoke
autoqa scan <url> --profile seo
autoqa scan --config autoqa.yaml
autoqa auth login <url>
autoqa setup
autoqa doctor
autoqa mcp

Documentation

• Project idea
• Architecture
• Usage design
• Authentication design
• Roadmap
• Agent instructions

Status

AutoQA is currently in the design stage. The first milestone is to make this command genuinely useful:

autoqa scan <url> --out <dir> --full

License

TBD.