In this video, I tested Claude Co‑work as a “no code, plain English” web scraper by pulling every sold eBay listing for a vintage card search over the last 30 days. I walk you through the exact setup: installing Claude Desktop, enabling Co‑work, and turning on the Claude in Chrome connector so Claude can actually control a browser tab (you’ll see the gold outline when it’s driving). Then I show the prompt I used, how I handled the annoying connection/proxy error, and what the output looked like when it worked: a clean spreadsheet with sold date, title, price, and sale type (auction vs buy it now), plus a summary tab. Before any scraping, I also cover the part most people skip: safety and rules. I show how to check a site’s robots.txt and remind you to read terms of service—because scraping the wrong site (LinkedIn is the classic example) can get you banned, and paywalled or sensitive sites can be blocked outright. Then I run a second demo on 130point.com, scraping two search terms and asking Claude to build a slide deck with the top sales and images. It worked, but it was slow and token-heavy (14% of my $100 plan for three pages), so my honest verdict is: Co‑work is great for one-off scrapes, but for repeated workflows you’re usually better off with Python, n8n, or third-party scrapers like Firecrawl.