Automate Any Website With Claude Code + Playwright CLI

Get AI income methods before they spread.

Free weekly intelligence newsletter.

Playwright CLI gives Claude Code full browser automation — clicking, form filling, screenshotting, and scraping — while keeping the AI context window lean. Three practical workflows show what this unlocks: automated QA with self-healing fix loops, Google Maps lead scraping, and an always-on community engagement bot.

Why Use Playwright CLI Instead of the MCP Server?

The MCP server version of Playwright loads all its tool definitions into Claude's context window every session. For a coding agent that already has your codebase, documentation, and task context loaded, that overhead adds up. The CLI version keeps context lean — Claude calls the browser as a subprocess rather than a persistent connected service.

Lean context means Claude stays focused longer before hitting limits.

For short automation tasks, the CLI is faster to set up and produces smaller, more portable scripts. The resulting automation becomes a reusable skill — a file you can call again for the same class of task without rebuilding it.

How Does the Automated QA Loop Work?

The QA loop runs in four steps and repeats until the page is clean:

Claude builds the feature or page
Playwright CLI takes a screenshot and checks for visual errors or broken elements
Claude reads the screenshot output and writes a fix
Playwright takes another screenshot — if clean, the loop exits; if not, repeat

The loop catches things that unit tests miss: layout breaks at specific viewport sizes, overlapping elements, missing assets, button states that only appear after interaction. It runs without a human reviewing each cycle.

For any project where visual correctness matters, this loop cuts QA time from hours to minutes.

How Does the Google Maps Scraping Work?

Claude Code with Playwright CLI navigates Google Maps, searches for a business category in a location, and extracts business names, phone numbers, addresses, and websites from the results. The script scrolls through paginated results automatically.

The output is a structured list of leads ready to import into a CRM or outreach tool. Nate ran this to build local business prospect lists — the kind of manual research that takes hours per city when done by hand.

One script, any city, any business category, running while you sleep.

What Is the Community Engagement Bot?

The bot monitors a community platform (Nate used School) for new posts, reads them with Claude, generates a relevant reply, and posts it — on a schedule. It is not spamming generic responses. Claude reads the actual post content and writes a contextual reply based on your defined persona and community role.

The practical use: staying present in a community you moderate or participate in without logging in manually every day. The bot maintains engagement between your actual appearances.

How Do You Set Up Playwright CLI With Claude Code?

Install Playwright CLI via npm: npm install -g @playwright/test and then playwright install chromium. In your Claude Code session, tell Claude to use Playwright CLI for browser tasks rather than asking it to open a browser manually. Claude will write the CLI commands as bash steps.

The first script Claude generates for any new automation task becomes a reusable template. Save it as a skill in your Claude Code skills folder so you can invoke it for similar tasks without rebuilding from scratch.

Frequently Asked Questions

Is Playwright CLI free to use?

Yes. Playwright is open source and maintained by Microsoft. The CLI and all browser engines (Chromium, Firefox, WebKit) are free to install and use for any purpose including commercial automation.

Does this work on headless servers?

Yes. Playwright runs in headless mode by default, which means it works on servers without a display. You can run browser automation on a VPS, a cloud VM, or any Linux environment. Add --headed if you want to watch the browser during development.

What is the difference between using this for QA versus scraping?

For QA, you are testing pages you own — checking that your code produces the right visual output. For scraping, you are extracting data from third-party sites. Both use the same Playwright commands. The difference is in what you do with the output and whether the site's terms of service permit scraping.

How do the scripts become reusable skills?

Save the Playwright script Claude generates as a .md file in your .claude/skills/ folder with a clear name like playwright-qa-loop.md. In future sessions, type /playwright-qa-loop and Claude executes it for the current project without you explaining how it works again.

Watch Nate demonstrate all three workflows: Nate Herk on YouTube