Point at a page.
Get the whole table.
Fyre turns any web page into clean, structured rows — product grids, search results, listings, reviews, directories, maps. It runs inside your browser, plugs into your AI agent, and scales up to scheduled runs when you need them.
- No code
- No subscription
- No data leaves your browser
Things people pull out of the web every day
Not a screenshot of a page.
A file someone else can use.
Every extraction ends in the same place: typed columns, real rows, and an export that keeps numbers numeric and dates parseable.
A product grid becomes a price list
Titles, prices, ratings, stock state, image and link — pulled from the grid you are looking at, with pagination followed to the end.
Three steps, and the second one is optional
There is no project to configure, no selector language to learn and no account required before your first table.
- 01
Click one thing
Open the side panel on any page and click a single example of what you want — one product card, one search result, one row of a directory.
- 02
Check the columns
Fyre finds every sibling and proposes a column for each field it can see, already named and typed. Drop what you do not need, re-point anything it got wrong.
- 03
Take the file
Rows land in a real spreadsheet in your browser. Dedupe, filter, merge with another run, then export CSV, Excel, JSON or push straight to a Google Sheet.
Same extractor. Three sets of hands.
Do it yourself in the side panel, hand it to an AI agent, or put it on a schedule. The recipe you build once works in all three.
You are already logged in. Use that.
Fyre lives in the Chrome side panel next to the page you are reading. It sees exactly what you see — the same session, the same cookies, the same rendered DOM after every script has run. There is no headless browser to fool and no proxy to configure, because the browser doing the work is yours.
Read more- Point and clickClick one item. Fyre infers the pattern, previews the columns, and waits for your approval before it touches anything.
- Nothing leaves the machineExtraction, storage and export all happen locally. Rows are written to IndexedDB in your profile and go nowhere else.
- Works behind a loginDashboards, member directories, internal admin panels — if the page renders for you, it extracts for you.
- No page limitThe licence is not metered. A ten-row table and a two-hundred-thousand-row table cost exactly the same: nothing more.
Give the agent hands, not a wall of HTML.
Fyre speaks the Model Context Protocol, so Claude, Cursor and any other MCP-capable client can drive it directly. The agent asks for a table; it gets rows. Raw markup never enters the conversation, which is the difference between a task that fits in context and one that does not.
Read more- Structured in, structured outThe agent receives typed columns and row counts, not a megabyte of div soup it has to squint at.
- Query before you readFilter, sort, aggregate and sample a hundred-thousand-row result server-side, then pull only the rows that matter into context.
- Thousands of pages per instructionOne sentence can start a run across an entire sitemap branch. Progress streams back as a count, not as content.
- Isolated workspacesEach agent session gets its own scratch space, so two parallel jobs cannot read or clobber each other's tables.
For the runs that should happen while you sleep.
Some jobs should not depend on a laptop being open. Promote any recipe you built in the extension to a cloud run: real browsers, parallel workers, a schedule, and delivery straight into the sheet or endpoint that consumes it.
Read more- Real browsers, not fetchPages render with JavaScript, lazy images resolve, and overlays are dismissed before extraction begins.
- Every ten minutes to once a monthPick an interval, a time zone and a page cap. Runs that find nothing new cost you nothing new.
- Delivered, not downloadedCreate, replace or append to a Google Sheet, or POST the batch to your own endpoint as JSON.
- Bring your own regionChoose where requests come from when a site serves different prices or stock by country.
Nine extractors, one shared table
Each tool does one job properly and writes into the same spreadsheet, so the output of one is the input of the next.
List Extractor
Repeating things become rows
Almost every useful page is a list wearing a costume — a product grid, a search results page, a directory, a leaderboard. Click one item and Fyre works out the pattern behind it, then pulls every sibling on the page into a table with sensible column names.
How it works- Rebuilt
Page Extractor
One template, a thousand detail pages
List pages rarely carry everything. Page Extractor takes a column of URLs — from a list you just built, a sitemap, or a pasted block — visits each one, and pulls the same set of fields from every page into a single table.
How it works Image Grabber
Every image, named the way you need it
Pull the images off a page — or off every page in a list — at full resolution, with filenames built from your own columns. No more folders of hash-named JPEGs you cannot match back to a product.
How it worksEmail Finder
Addresses out of the noise
Sweep a page, a list of pages, or a whole sitemap branch for email addresses. Fyre reads the places addresses actually hide — mailto links, obfuscated text, contact blocks, structured data — then dedupes what it finds.
How it works- New
Shopify Extractor
A whole storefront, variants and all
Shopify stores expose a clean product feed if you know where to look. Fyre reads it directly, so you get the full catalogue — every variant, price, SKU, option and image — without clicking through a single collection page.
How it works Social Link Extractor
Profiles, separated and labelled
Collect the social presence of every business on a page or in a list. Fyre recognises the major networks, normalises the URLs, strips tracking parameters, and gives each network its own column instead of one messy pile of links.
How it worksText Extractor
The article, without the furniture
Turn a page into clean reading text or Markdown with the navigation, cookie bars, related-posts rails and newsletter interruptions removed. Useful for research, and the cheapest way to feed a page to a language model.
How it works- New
Sitemap Explorer
The site's own map of itself
Most sites tell you exactly what they contain — you just have to read the sitemap. Fyre resolves nested index files, expands every branch into a browsable tree, and lets you filter down to the section you want before handing the URLs to another tool.
How it works Data Table
Where everything lands
Every tool writes into the same place: a real spreadsheet in a browser tab, backed by local storage that survives a crash. Sort, filter, dedupe, edit, merge two runs, then export — or hand a column to the next tool without leaving the page.
How it works
Or start from a recipe
Saved templates for the sites people ask about most. Same extractors underneath, with the guesswork already done.
Map results scraper
Map search results
Marketplace product scraper
Large retail marketplaces
Review platform scraper
Public review platforms
Business directory scraper
Local and trade directories
Job board scraper
Job boards and careers pages
Property listing scraper
Property portals
Social profile scraper
Public social profiles
Storefront catalogue scraper
Hosted storefronts
Fifteen dollars.
That is the whole pricing page.
Scraping tools have a habit of charging monthly for something that runs on your own machine. Fyre does not. You pay once, you own the licence, and the row counter that every competitor uses to price you simply does not exist here.
- No monthly fee and no annual renewal
- No per-row, per-page or per-credit metering
- No seat pricing — buy one licence per person
- Updates included for as long as the product exists
One payment unlocks everything below on up to three machines. No seats, no row quota, no renewal date sitting in your calendar.
- All nine extractors, no tool held back
- Unlimited rows, pages and extractions
- Pagination, infinite scroll and load-more handling
- The full data table: dedupe, merge, filter, edit
- CSV, Excel, JSON, Google Sheets and webhook export
- MCP server for Claude, Cursor and other agents
Free tier available first — 200 rows per extraction, every tool unlocked.
What people say
Collected from the Chrome Web Store and from email, with permission.
Placeholder testimonial. Replace with a real review before publishing. Roughly two sentences works best in this card — long enough to say something specific, short enough to scan.
Sample reviewerE-commerce analyst Placeholder testimonial. Replace with a real review before publishing. Reviews that name the actual task tend to convert better than reviews that only say the product is good.
Sample reviewerAgency founder Placeholder testimonial. Replace with a real review before publishing. Aim for a mix of use cases across the set: lists, detail pages, images, agent workflows.
Sample reviewerMarket researcher Placeholder testimonial. Replace with a real review before publishing. Including at least one four-star review reads as more credible than a wall of fives.
Sample reviewerData engineer Placeholder testimonial. Replace with a real review before publishing. Keep the role label accurate — it does more work than the name does.
Sample reviewerRecruiter Placeholder testimonial. Replace with a real review before publishing. If you have permission to link to the reviewer, add a `url` field and render it here.
Sample reviewerReal estate consultant
Still being built
A short log of what shipped over the last few months.
- Featurev2.4.0 · 2026-08-14
Page metadata as first-class columns
Page Extractor can now capture title, meta description, canonical, Open Graph, Twitter card and JSON-LD without a manual selector for each.
- Improvementv2.3.2 · 2026-08-02
Filename templates in Image Grabber
Filenames can now be composed from any combination of table columns, plus {index} and {ext}.
- Featurev2.3.0 · 2026-07-21
Page Extractor, rebuilt
Rewritten around a worker pool: concurrency and delay are now separate settings you control per run.
- Featurev2.2.0 · 2026-07-08
Sitemap Explorer
New tool: resolve a site's sitemap, follow index files, and browse the result as a path tree with counts.
The things people ask before they install
If yours is not here, the full list has eighteen more — and support answers email from a real inbox.
It turns web pages into spreadsheets. You install a Chrome extension, open the side panel next to any page, click one example of the thing you want, and Fyre works out the repeating pattern and pulls every match into a table. From there you filter it, dedupe it, and export CSV, Excel, JSON, or push it to a Google Sheet.
No. The whole flow is clicking on the page and adjusting columns. If you do write code, CSS selectors and regular expressions are available wherever a column is defined — but they are an escape hatch, not the main path.
Yes. One payment of $15 unlocks every extractor, every export format, the AI agent connection and future updates. There is no monthly plan, no per-seat pricing, no row quota and no upsell waiting three screens in.
Into your own browser. Rows are written to IndexedDB inside your Chrome profile, and they stay there until you export or delete them. Local extraction involves no upload of page content or results to us. Scheduled cloud runs are the exception — those execute on our infrastructure by definition, and the page explains exactly what is stored and for how long.
That is exactly the problem it is built around. Results live in a table on Fyre's side; the agent gets a schema, a row count and a handle. It then filters, aggregates or samples that table and pulls in only the rows it needs. A 120,000-row run can cost a few hundred tokens in conversation.
The recipe will start returning empty columns, and Fyre tells you that instead of writing blank rows quietly. Re-pointing a column takes one click on the new element. For scheduled runs you can set an alert threshold so an empty result emails you rather than overwriting yesterday's good data.
Stop copying tables by hand.
Install the extension, click one row, and watch the rest arrive. If it does not do what you need in the first ten minutes, you have lost nothing.