mikeparcewski/agent-browser

Browser Automation

Detects and uses the best available browser automation CLI in your environment.

CLI Discovery

Before executing browser automation, detect which tool is available:

# Priority order: playwright > puppeteer > cypress > selenium-webdriver
for tool in playwright puppeteer cypress selenium-webdriver; do
  if command -v "$tool" > /dev/null 2>&1; then
    echo "Using: $tool"
    break
  fi
  # Check via npx for Node-based tools
  if command -v npx > /dev/null 2>&1; then
    if npx "$tool" --version > /dev/null 2>&1; then
      echo "Using: npx $tool"
      break
    fi
  fi
done

Stake level: Medium — inform the user which tool was selected.

If none found: Recommend Playwright installation (see below).

Pattern: This follows the integration-discovery CLI detection pattern. See integration-discovery refs/cli-detection.md for the full decision policy.

Priority Order

| Priority | Tool | Why | |----------|------|-----| | 1 | Playwright | Best AI-agent support, cross-browser, built-in waiting | | 2 | Puppeteer | Mature API, Chrome-focused, wide adoption | | 3 | Cypress | Test-focused, good DX, component testing | | 4 | Selenium WebDriver | Legacy support, multi-language |

Installation (None Found)

# Playwright (recommended)
npm i -D playwright
npx playwright install

# Or with uv (Python)
uv add playwright
python -m playwright install

Quick Command Reference

| Scenario | Playwright | Puppeteer | Cypress | |----------|------------|-----------|---------| | Run tests | npx playwright test | node script | npx cypress run | | Open UI | npx playwright test --ui | - | npx cypress open | | Record test | npx playwright codegen URL | - | - | | Screenshot | page.screenshot() | page.screenshot() | cy.screenshot() | | Navigate | page.goto(URL) | page.goto(URL) | cy.visit(URL) | | Click | page.click(sel) | page.click(sel) | cy.get(sel).click() | | Fill input | page.fill(sel, val) | page.type(sel, val) | cy.get(sel).type(val) | | Wait | page.waitForSelector() | page.waitForSelector() | built-in |

Common Patterns

Screenshot a Page

# Playwright
npx playwright screenshot --browser chromium https://example.com out.png

Run Existing Tests

# Playwright
npx playwright test

# Cypress
npx cypress run

# Puppeteer (node script)
node tests/puppeteer.js

Record a New Test (Playwright only)

npx playwright codegen https://example.com
# Opens browser + code recorder, paste generated test into your test file

Accessibility Audit

# Playwright + axe-core
npx playwright test --grep @a11y

# agent-browser CLI (legacy)
agent-browser open https://example.com
agent-browser a11y

Legacy: agent-browser CLI

If agent-browser is installed (older projects), it provides a simpler command interface:

agent-browser open https://example.com
agent-browser screenshot --output page.png
agent-browser snapshot    # accessibility tree + content
agent-browser click "button.submit"
agent-browser fill "input[name=email]" "[email protected]"
agent-browser a11y        # accessibility violations

References

• Playwright - Full Playwright usage patterns
• Puppeteer - Puppeteer API and patterns
• Cypress - Cypress test runner guide
• Selenium - Selenium WebDriver patterns