Skip to content
The extension

The unglamorous details that decide
whether a scrape actually works

Anyone can grab the first twelve rows of a list. The difference between a demo and a tool is what happens on page nineteen, when the layout changes, the images are lazy-loaded, and half the prices are inside a currency span.

Selection that gets it right

The part everything else depends on.

  • Pattern inference from one click

    Pick a single item and Fyre works backwards to the repeating structure, then shows you every sibling it matched so you can correct it before extracting.

  • Automatic column naming

    Columns arrive as price, title, rating and link — read from labels, microdata and position — rather than as div_3 and span_7.

  • Type inference

    Prices become numbers with the currency kept aside, dates parse to ISO, ratings become floats, and relative links resolve to absolute URLs.

  • Manual override anywhere

    Any column can be re-pointed at a different element, given a fixed value, or defined by a CSS selector you write yourself.

Getting through the whole list

Pagination is where most scrapers quietly give up.

  • Numbered pages

    Fyre finds the next control, follows it, and stops at the page cap you set or when the rows stop changing.

  • Infinite scroll

    Scroll is driven at a human pace with a configurable idle threshold, so lazy loaders actually finish before the next step.

  • Load-more buttons

    Detected and clicked repeatedly, with a guard against the button that quietly stops doing anything after page forty.

  • Resume where it stopped

    Close the tab mid-run and the partial table is still there. Restart from the last completed page instead of the beginning.

Data you can actually hand over

A scrape is only finished when someone else can use the file.

  • Deduplication that explains itself

    Choose the key columns and preview exactly which rows collapse into which before anything is removed.

  • Merge across runs

    Join a list-page table to a detail-page table on a shared URL or SKU and get one wide table out.

  • Inline editing

    Fix the four rows the site got wrong without exporting, opening a spreadsheet, and re-importing.

  • Exports that keep types

    CSV, Excel, JSON, Google Sheets or a webhook — with numbers still numbers and dates still dates on the other side.

Built to stay out of the way

Speed, privacy and the boring reliability details.

  • Local by default

    No account is required to extract. Rows live in your browser profile, and clearing them is a single button in Settings.

  • Speed profiles

    Gentle, normal or fast, controlling concurrency and delay. Gentle is the default because it is the one that does not get you blocked.

  • Honest failure

    Pages that fail are surfaced with the reason and the URL, not silently dropped to make the count look better.

  • Keyboard-first

    Every step of the flow has a shortcut, and the picker can be driven with arrow keys when the mouse is in the way.

The toolkit

Nine extractors, one shared table

Every tool writes into the same spreadsheet, so a column from one becomes the input of the next without an export in between.

$15 · one payment · lifetime

Everything above, for fifteen dollars.

One payment, three machines, every future update. The free tier lets you prove it works on your pages first.