Most free SEO tools ask you to paste something in. That is fine for a single check and useless the moment you are looking after ten sites, because the thing you want to know is almost never about text you already have in your clipboard. It is about a page that is live right now, on a domain you did not build, doing something you did not expect.
So these take a URL. The robots.txt tester fetches the live file and tells you which crawlers it actually lets through. The meta tag checker takes fifty URLs and returns the titles that are too long in pixels rather than characters. The redirect checker traces a hundred URLs through their chains at once. Every one of them exports to CSV, because the table is usually the beginning of the work rather than the end of it.
The tools that need judgment rather than a fetch use a language model: clustering a keyword list into the articles that should exist, turning a subject into a topical map, or writing five meta descriptions for a page it has read. Those are the parts of SEO that used to need an afternoon and a spreadsheet.