Every job below names the tool that does it. Where we have a capture of that job, it is a real screenshot of the extension running in Chrome rather than a drawing. If yours is not here, it is probably one of these with a different site.
Company and contact lists
Turn a company directory into a call list
Two hundred companies, spread over ten pages, and the sheet is due today.
Click one entry. Keep the columns you care about, set a page cap and let it walk the rest: one row per company, with the site, the branch and whatever contact fields the listing shows.
This job collects data about people. You decide the legal basis for holding it and you answer for that decision. See the terms
Take the exhibitor list before the fair starts
The exhibitor catalogue is a web page with four hundred names on it. You have two days at the fair and would like to plan them.
One click on an exhibitor card takes the whole catalogue: stand number, sector and site. Sort it in your spreadsheet and you have a route for both days.
Their shop shows twenty products per page. You want the whole assortment next to yours, in one view.
Click one product card and Tabitha finds the other nineteen. It proposes columns for name, price, availability, image and link, then follows the next button to the end of the catalogue.
You need to know when they cut a price. Tools that do this start at a monthly fee per shop.
Save the run as a recipe and start it once a week. Put two exports next to each other in a spreadsheet and you have a price history nobody had to type.
Four hundred reviews sit behind a load more button. What you want is the thing people repeat, and the first ten will not tell you.
Author, score, date and the full text each get a column. The score comes back as a number, so you can average it yourself or hand the whole file to Claude.
Your search returns sixty flats and the good ones go in two days. The portal will not export any of them.
Take price, area, district, floor and link into a table. Run the same recipe tomorrow, and the difference between the two files is what came on the market.
Somebody asks how many players there are in a niche. The answer is scattered across directories and rankings.
Extract each source into the same table and deduplicate by domain. Every row keeps its link, which is what makes the count defensible when somebody questions it.
You know what you want out of the page and you would rather describe it once than set the columns yourself.
Connect the MCP server and say what to collect. Claude drives the same tools in the browser you already have open, and what comes back to the chat is rows.
Six hundred pages, and somewhere in them duplicate titles and empty descriptions. One tab at a time will never show you which.
Feed in the address list and take title, description, canonical, first heading and the Open Graph fields. Sorting by title puts the duplicates next to each other.