Free tool + Chrome extension · v2.1.0 live on the Web Store
Copy any table on any web page into CSV, Excel, Markdown, JSON, SQL and eight more formats. Click the one you want, drag a box round something that is not a table at all, or paste your own data. Merged cells stay aligned, numbers stay numbers, and every row comes with you.
Select a table on any page, copy it, and paste it below. Browsers put the real HTML on the clipboard, so the structure survives — merged cells and all. This runs the same engine as the extension, entirely in your browser.
Copy a table from any web page and paste it here — browsers put the real HTML on the clipboard, so merged cells survive. CSV, TSV, JSON, JSON Lines and Markdown tables work too, and the format is detected for you.
With the extension installed, the page itself becomes the input. Every one of these ends the same way — a panel telling you what was found and how it was read, with copy buttons and a way back, including the attempts that come up empty.
Click the toolbar icon and every table on the page is listed with its row and column count. Copy one straight from the list as CSV, TSV or Markdown, or open it in the editor.
Alt+Shift+TPoint at it. The outline follows the cursor with a live row and column count, and ↑ / ↓ widen or narrow the selection when it lands on the wrong container. Any element will do — if it is not a table, it is turned into one.
Alt+Shift+GFor the things that only look like tables. The text inside the box is clustered back into rows and columns by where it sits on screen. Cover half a real table and both readings are offered, with counts — the part you covered, or the whole thing.
Alt+Shift+BPaste CSV, TSV, JSON, JSON Lines, Markdown or an HTML table into the editor, or drop a file onto it. The format is detected and the delimiter sniffed, so it converts in every direction rather than only out of a page.
Paste or dropCopying a table sounds trivial until you meet a real one. These are the 7 cases that break ordinary extractors, and what Table Grabber does about each.
One cell spanning two rows occupies several positions in the grid. Read the cells in order and every column after the merge shifts left — so the numbers end up under the wrong headings, and nothing warns you.
What it does: Every cell is laid onto a coordinate grid before anything is read, so a merged cell still covers the positions it visually covers. You choose how those positions are filled: repeat the value, leave them blank, or mark them.
A grouped header — one 'Region' cell sitting above 'Country' and 'City' — is two header rows. Take only the first and you lose the columns; take only the second and you lose the group.
What it does: Header rows are detected, then stacked into one name per column, so you get 'Region – Country' rather than half the information.
Plenty of web apps draw grids out of plain boxes with CSS. There is no table markup at all, so a tool looking for <table> finds nothing on a page that visibly contains a table.
What it does: Repeating structure is the signal: a container whose children look alike and each hold the same number of cells. That finds hand-rolled grids and the big JavaScript grid libraries alike, without hardcoding any library's class names.
A price list in flexbox, a wall of dashboard tiles, a receipt in a <pre> block: you can see the rows and columns, but they exist nowhere in the markup — only in where the text landed on screen. No amount of reading the page finds a table, because there is not one.
What it does: Drag a box round it and the text inside is clustered back into a grid by position: fragments sharing a horizontal band become a row, fragments sharing a vertical band become a column. Right-aligned number columns, uneven row heights and missing values are all handled.
Large grids keep only the visible rows in the page and swap them as you scroll. Read the page once and you get twenty rows out of fifty thousand — and the twenty look perfectly complete.
What it does: The grid is scrolled in steps and the rows accumulated, re-read at each step with whichever reader matches the grid — so a grid built from divs collects too, not just a real table. Where the grid publishes its own row numbers, those place each row exactly; otherwise rows are matched by content, and you are told which happened.
Pagination hides most of the data behind a Next button, and copying ten pages by hand is how an afternoon disappears.
What it does: The next control is found, clicked, and the new rows are collected once the table has actually changed — not after a hopeful fixed delay. Every page is merged into one export. It stops on the last page, and you can stop it sooner.
A table inside a shadow root or an iframe is invisible to an ordinary search of the page, so tools report there is nothing there.
What it does: Open shadow roots and same-origin iframes are searched too. Cross-origin frames are skipped, because the browser correctly forbids reading them and no extension should try.
Marked formats are ones the most-installed table extension charges a subscription for. They are all pure computation in your browser, which is why they are free here.
Comma-separated, RFC 4180 quoting, UTF-8 BOM so Excel opens accents correctly.
Tab-separated. Pastes into Sheets and Excel as real cells, not one.
A real workbook with numeric cells, a frozen header and sized columns.
Array of objects keyed by header, with real numbers and booleans.
Array of arrays, header row first. Closer to the original grid.
One record per line — what log pipelines and bulk loaders expect.
One element per row, with header names sanitised into legal tag names.
GitHub-flavoured, padded to align, numeric columns right-aligned.
A clean table with none of the original page's styling.
CREATE TABLE with inferred types, plus batched INSERTs. Postgres, MySQL or SQLite.
One document, a list of records, correctly quoted.
A tabular environment with the special characters escaped.
Fixed-width columns, aligned. For terminals and code comments.
Getting the rows out is the easy half. Most of the work went into the data being usable when it arrives.
If a page has fifteen tables, pick the one you mean by clicking it, with ↑ and ↓ to widen or narrow the selection. The overlay is drawn inside a shadow root, so it looks the same on every site no matter how aggressive the page's own CSS is — and Cancel, Esc or a right-click always gets you out.
Drag a box over a flexbox price list, a set of stat tiles or a block of fixed-width text and it is rebuilt into rows and columns from where the text sits on screen. Cover part of a real table and you are offered both readings with counts, rather than one being guessed at silently.
Paste CSV, TSV, JSON, JSON Lines or a Markdown table into the editor, or drop a file onto it, and it reads those too — guessing the format and sniffing the delimiter. So it converts in every direction: CSV to Markdown, JSON to a spreadsheet, a web table to SQL.
Double-click any cell and type. Enter commits and drops a row, Tab moves right, Escape abandons it. The value goes in underneath the clean-up settings rather than on top of them, so it is treated exactly as though the page had said it — and it is what gets exported.
Drag a row by its number, or move the selected one with Alt and an arrow key. Insert a blank row, duplicate one, delete one. Rearranging while a sort is running freezes that sort into a real order first and tells you — a row moved under a live sort would only spring back.
Undo and redo run a hundred steps deep across edits, moves, insertions, deletions, hides and renames. Every change is a layer on top of the grabbed table rather than a write into it, so one button returns every value to what the page actually said.
Reorder columns by dragging their headers in the grid or the side panel, rename or hide them, transpose the whole thing, fill blank cells from the value above, find and replace across every cell, trim whitespace and footnote markers, and rewrite headers to snake_case or camelCase to match wherever the data is going.
Click a header to sort: numbers by value so 9 comes before 100, dates chronologically rather than alphabetically, blanks last in both directions, and ties left in the order they arrived. The sort is pinned to the column you chose, so hiding or reordering the others does not move it.
Every column reports its detected type, how many cells are filled, how many values are distinct, and for numbers the min, max, sum and mean. It is how you notice a capture came up short or an ID column has duplicates.
$1,234.00 pasted into a spreadsheet is text, and your SUM() returns zero. Column types are detected and values can be rewritten so the destination reads them as numbers — including European 1.234,56 and accounting negatives like (1,234).
The .xlsx export writes real numeric cells, a frozen header row and sized columns, so formulas and charts work the moment the file opens. No import wizard, no text-to-columns.
Copying as TSV also puts an HTML table on the clipboard, so Google Sheets and Excel receive real cells instead of one long string in A1. The copy buttons sit on the in-page result panel too, so a quick grab never has to open the editor.
A scraped cell reading =cmd|'/c calc'!A1 is a live formula the moment your CSV opens in Excel. Values starting with = + - @ are neutralised on export, which matters when the source is a page you do not control.
Checked in August 2026 against the published listings and pricing pages of the most-installed table extensions. These tools are good; the point is narrower — the features behind their paywalls are pure browser-side computation, with no server and so no running cost.
| Capability | Typical paid extension | Table Grabber |
|---|---|---|
| Rows per export | Commonly capped at 250 on the free tier | Unlimited |
| Copy as Markdown | Typically a paid feature | Free |
| Download CSV / Excel directly | Typically a paid feature | Free |
| Multi-page and infinite-scroll capture | Typically a paid feature | Free |
| Real .xlsx with numeric cells | Often an HTML file renamed .xls | A real workbook |
| Merged cell handling | Varies; several misalign columns silently | Grid-accurate, with a choice of fill policy |
| Editing cells before export | Rare; usually export-only | Type into any cell, with undo |
| Reordering rows before export | Sorting only | Drag rows, insert, duplicate, delete |
| Tables built from divs | Sometimes supported | Found by repeating structure |
| Capturing an area that is not a table at all | Rare | Drag a box; rebuilt from screen positions |
| Shadow DOM and iframes | Rarely supported | Searched |
| SQL, YAML and LaTeX output | Rare | Included |
| Site access required to install | Often all sites, always on | None at all |
| Account required | Sometimes | Never |
| Price | $12–$120 a year | Free, and open source |
These are the tools people are usually choosing between, and where each one stops. Checked in August 2026 against their own store listings and pricing pages.
The most-installed table extension, and the one most people mean by “the table one”.
The usual free recommendation, and a genuinely good one.
A lightweight one-job extension.
A general-purpose scraper that guesses at repeating structures across pages.
Select, Ctrl+C, paste, then fix it.
Pull a reference table off a stats site into Excel with the numbers already numeric, instead of retyping it or fighting text-to-columns.
Copy a table straight into a README or a blog post as Markdown, aligned and correctly escaped, in one click.
Turn a table of reference data into a SQL seed script, a JSON fixture or a typed array without writing a parser for a one-off job.
Get a LaTeX tabular out of a web table without rebuilding it by hand, footnote markers already stripped.
Published to the Chrome Web Store on 26 August 2026. Until now the editor could reshape a table — hide a column, sort it, rename a header — but it could not change a single value, so one typo at the source or one row in the wrong place sent you to a spreadsheet to finish by hand. Now the grid itself is editable and the row order is yours. If you have an earlier version installed, Chrome will update you on its own.
Double-click a cell, or select it and start typing. Enter commits and drops to the next row, Tab moves right, Escape abandons the change. Whatever you type is what gets exported — no round trip through a spreadsheet to fix one typo.
Drag a row by its number, or move the selected one with Alt and an arrow key. Insert a blank row, duplicate one, delete one. Rearranging while a sort is running freezes that sort into a real order first and says so, because a row moved under a live sort would only spring back.
Drag a header to reorder it, or use the pencil on it to rename in place — no trip to the side panel. Hidden columns keep their position in the order, so unhiding one later puts it back where it belongs rather than at the end.
Ctrl+Z and Ctrl+Shift+Z cover edits, moves, insertions, deletions, hides and renames. The side panel keeps a running count of what you have changed, with one button that discards all of it.
Edits are a layer keyed by row identity rather than by position, so a change follows its row through a sort, a filter and a reorder — and the table as the page served it is always one click away.
Arrow keys, Home, End, PageUp and PageDown move the selection; F2 opens a cell; Alt with an arrow moves the row or column itself; Alt+Enter inserts a row; Ctrl+Delete removes one.
The web version above handles anything you can copy. The extension does the three things a web page is not allowed to: read a table in place, scroll a grid that only loads rows as you go, and walk every page of a paginated table.
The Add to Chrome button goes straight to it. Click Add to Chrome there and confirm. Every Chromium browser installs from the same listing — Edge, Brave, Opera, Arc and Vivaldi included, though Edge shows a one-time banner asking you to allow extensions from other stores.
Click the puzzle-piece icon in the toolbar and pin Table Grabber so it is one click away.
Open any page with a table and click the icon, or press Alt+Shift+T. To point at a particular one, press Alt+Shift+G and click it; to capture something that is not a table, press Alt+Shift+B and drag a box round it.
Byte for byte the same build as the listing, installed the way Chrome tests every extension before it is published. You give up automatic updates and gain the ability to read every line before you run it.
Unzip it somewhere you will keep — Chrome loads the extension from this folder every time it starts, so a temporary folder will break it later.
Paste this into the address bar and press Enter.
chrome://extensionsThe switch is in the top-right corner of that page.
Click Load unpacked and choose the table-grabber folder inside the archive — the one containing manifest.json.
Pin it from the puzzle-piece menu, then open any page with a table and press Alt+Shift+T.
Installing unpacked means trusting a file from the internet, so here is its fingerprint. Run one of these on the ZIP and compare — if the hash matches, the file arrived exactly as it was built.
c4fa5c052074042af6f8f41c5a1deb424c366ff9b1d872cf8a09b2b7f8b1f883Get-FileHash table-grabber-2.1.0.zipshasum -a 256 table-grabber-2.1.0.zipsha256sum table-grabber-2.1.0.zipThe short version: nothing is collected, because there is nothing in it capable of collecting anything.
This exists because of a one-star review. Someone had used a popular table extension for years, recommended it to everyone they knew, and then found the free tier limited to 250 rows. Another reviewer put the objection more precisely: needing a subscription to copy as Markdown is odd, because converting a table to Markdown requires no backend, and therefore has no continuous operating cost to recover.
That is exactly right, and it generalises. Reading a table out of a page, expanding its merged cells, working out which columns are numbers and writing the result as CSV, Markdown or a spreadsheet is arithmetic. It happens on your own machine, in your own browser, on data your browser already downloaded. Nobody is paying for servers to make it happen, because there are none.
So the whole thing is free and it will stay free, which is a promise that costs nothing to keep. There is no account to create, no row limit to hit, no format held back, and no upgrade screen. It is MIT licensed and the source is public, so you do not have to take any of that on faith.
The privacy position follows from the same logic. A tool that does its work locally has no reason to make a single network request, so this one contains no network code at all — and that is checked automatically, with the build failing if a fetch ever appears in the source. It installs with access to no websites, and reads a page only in the moment you ask it to.
Yes. As of August 2026 Table Capture's published pricing caps free exports at 250 rows, with PRO Basic at $12/year for 1,000 rows and PRO Max at $120/year for unlimited. Table Grabber has no row limit and no paid tier at all — a 40,000-row table exports exactly like a 12-row one. It also adds SQL, YAML, LaTeX and TSV output, handles tables built from divs, and can capture an area of a page that is not a table. Table Capture is the more established tool with a longer track record and a more polished Google Sheets round-trip; if you are hitting the free row cap, this is the free way past it.
It depends on what breaks for you. For a plain HTML table, almost any of them work — Table Capture, HTML Table Exporter and Table to Excel all do the simple case. What separates them is the awkward cases: row limits on the free tier, tables built from divs rather than <table>, merged cells that silently misalign columns, rows that only load as you scroll, and content inside a shadow DOM or an iframe. Table Grabber handles all of those without a paid tier, and writes a real .xlsx workbook with numeric cells rather than an HTML file renamed .xls.
Row caps are a pricing tier, not a technical limit — the work happens in your browser either way. Table Grabber does not have one: use it here on this page or install the extension, and export the whole table however many rows it has. If the rows only appear as you scroll, turn on scroll-capture and it will collect them as they load rather than exporting only the visible window.
Really free, permanently, and open source under the MIT licence. The reasoning is not generosity: everything it does happens in your browser, so it costs nothing to run and there is nothing to recover. A row limit would exist only to sell you the absence of a row limit.
Yes, and it is the main addition in 2.1. Double-click any cell in the editor and type, or select it and start typing. Enter commits and drops to the next row, Tab moves right, Escape abandons the change. Whatever you type is what gets exported, so a typo at the source or a value the page got wrong no longer means a round trip through a spreadsheet. Undo runs a hundred steps deep, and a single button in the side panel puts every value back as the page had it — your edits sit in a layer on top of the grabbed table rather than being written into it.
Drag a row by its number in the editor's grid, or select a cell in it and press Alt+↑ or Alt+↓. You can also insert a blank row, duplicate a row and delete a row from the toolbar above the grid. If a sort is running when you move something, Table Grabber freezes that sort into a real order first and tells you it did — otherwise the row you moved would simply spring back to wherever the sort wants it.
Drag a column header sideways in the grid, or drag its entry in the Columns list on the left; the two stay in step. The pencil on a header renames it in place, and so does clicking the name in the side panel. Hidden columns keep their position in the order, so unhiding one later puts it back where it belongs rather than at the end. Alt+← and Alt+→ move the selected column without a mouse.
No. The editor is a separate tab and it never writes to the page it read — that was true before and it is still true now that you can type into it. Edits live in the editor tab's memory, are applied on top of the grabbed copy, and vanish when you close the tab. Nothing is saved to disk except the file you explicitly download.
Transposing swaps rows and columns, so what was a row is now a column and there is no stable row identity left for an edit to attach to. Rather than guess which cell you meant and write to the wrong one, the grid goes read-only and says so. Untick Transpose, make your edits, then transpose again — the order is preserved either way.
No. The editor draws the first few hundred rows for speed, and says so where you can see it, but exports always contain every row.
No, and it cannot be. The extension contains no network code — no fetch, no XMLHttpRequest, no WebSocket, no analytics. This is enforced by a check in the build that fails if any of those appear in the source, so it is verifiable rather than merely promised.
It installs with access to no websites at all — there is no host permission in its manifest, so nothing runs anywhere until you invoke it. When you click the icon it reads that one tab. It also asks for storage (to pass a table to the editor tab, in memory), downloads (to save your file) and contextMenus (for the right-click items). Neither 2.0 nor 2.1 added a permission — the editable grid in 2.1 needs none, because everything it does happens in the editor tab you are already looking at.
Yes. Press Alt+Shift+B, or choose 'Drag a box', and draw a rectangle round whatever you want — a flexbox price list, a grid of dashboard tiles, a receipt in fixed-width text. The text inside is clustered back into rows and columns by where it sits on screen, so it comes out as a table even though the page never contained one.
Press Esc, right-click, or use the Cancel button on the control bar; reopening the popup mid-pick offers a Cancel too. Clicking a table that lives inside a link no longer follows the link, so a mis-click cannot navigate away and lose the grab.
Alt+Shift+T opens the popup and lists every table on the page, Alt+Shift+G starts a click-pick, and Alt+Shift+B a box-drag. While picking, the arrow keys widen or narrow the selection and Enter takes it. In the editor, click a column header to sort, double-click a cell (or press F2) to edit it, the arrow keys move the selection, Alt with an arrow moves the selected row or column, Alt+Enter inserts a row, Ctrl/Cmd+Delete removes one, Ctrl/Cmd+Z undoes, Ctrl/Cmd+S downloads, Ctrl/Cmd+Shift+C copies and Ctrl/Cmd+F jumps to the row filter. All three browser shortcuts can be rebound at chrome://extensions/shortcuts.
That is the merged-cell problem, and it is the main thing Table Grabber is built to get right. If you are seeing it here rather than in another tool, the table has a structure worth reporting — please send the URL.
That grid keeps only the visible rows loaded. Open the editor and use 'Collect all rows', which scrolls the grid and accumulates the rest — including grids built out of plain boxes. If the data is paginated instead, use 'Walk every page'. Do it before you start editing: collecting re-reads the table from the page and renumbers its rows, so it asks first and clears any edits already made.
Because the page wrote them as $1,234.00, which is a string. Tick 'Values as numbers' in the editor, or export as .xlsx, which writes genuine numeric cells so formulas work immediately.
Yes. The current version, 2.1.0, went live on 26 August 2026, so the normal way to install it is the Add to Chrome button at the top of this page — one click, and Chrome keeps it updated from then on.
Yes, and it is the reason for the rebuild. The pick was being answered into a popup that had already closed, so the button really was dead. The page now completes the grab itself, and every attempt ends in a panel that says what happened — including the ones that fail. If Chrome has not updated you yet, open chrome://extensions and press Update; you want 2.1.0 or newer.
The Web Store, unless you have a reason not to: it is one click and it updates itself. The ZIP is the same build, kept here for anyone who would rather read every line before running it, or who cannot install from the store — you load it unpacked, and a published SHA-256 on this page lets you check the file arrived intact. Unpacked extensions do not update themselves.
Yes. The web version on this page runs the same extraction and export engine — copy a table from any page and paste it in, or paste the HTML. The extension exists for the things a web page cannot do: reading a table in place, dragging a box over part of a page, scrolling a virtualised grid, and walking pagination. Cell editing and row rearranging are in the extension’s editor.
Not yet. It is built for Chromium browsers — Chrome, Edge, Brave, Opera, Arc and Vivaldi, version 116 or newer. A Firefox port is plausible since the code is nearly all standard web APIs.
It reads what is on screen, which is the point, so an exported table can contain personal or account data — treat the file accordingly. Nothing is transmitted, and access ends when you leave the page.
Something not working? Start here — these 11 cover nearly everything. If none of them fit, the report link at the bottom goes straight to me.
Why: What is on screen may not be a table at all — a layout of boxes with no repeating structure to recognise — or it lives in a cross-origin iframe, which no extension is permitted to read.
Fix: Use “Click a table” and click it directly; that reads whatever you point at. If it still comes out wrong, use “Drag a box” and draw one round it, which rebuilds a table from where the text sits rather than from the markup. For an embedded frame from another domain, open that frame in its own tab first.
Why: The grid keeps just the visible rows loaded and swaps them as you scroll, so a single read genuinely only sees those.
Fix: Open the editor and press “Collect all rows”. If the data is paginated instead, use “Walk every page”. Leave the tab visible while it runs; a scroller that refuses to move now stops early and says so rather than grinding.
Why: A box that overlaps a real table is ambiguous: you might mean the part you covered, or the table it belongs to.
Fix: Both readings are offered with their row and column counts — take the other one. For a layout with no table behind it, redraw the box slightly wider: columns are found from the vertical bands the text falls into, so clipping a column edge loses it.
Why: Almost always merged cells. This is the case Table Grabber is specifically built to get right, so seeing it here is a bug worth reporting.
Fix: Please send the page address — a table that defeats the grid builder is the single most useful thing you can report.
Why: The page wrote them as $1,234.00, which is a string, not a number.
Fix: Tick “Values as numbers” in the editor, or export as Excel (.xlsx) — that writes genuine numeric cells, so formulas work the moment the file opens.
Why: A sort was running. A sort is recomputed from the data every time, so it has the last word over any order you set by hand.
Fix: Nothing to do — the first move you make freezes that sort into a real order and tells you it has, and every move after it sticks. To choose the moment yourself, clear the sort first by clicking the chip beside the row filter.
Why: Either Transpose is on, or the cell is in a row the row filter is currently hiding.
Fix: Untick Transpose in the Shape panel — swapping rows for columns leaves no stable row for an edit to attach to, so the grid goes read-only rather than writing to a cell you did not mean. Otherwise clear the filter box, edit, and re-apply it.
Why: “Collect all rows” and “Walk every page” re-read the table from the page, which renumbers its rows.
Fix: Collect first, then edit. The editor asks before running either of them when edits exist, precisely because reapplying an edit to whatever row now holds that number would be worse than losing it.
Why: A selection is still running.
Fix: Press Esc, right-click anywhere, click Cancel on the control bar, or reopen the popup and press Cancel there. Any of the four ends it and leaves the page untouched.
Why: Chrome does not allow extensions to run on its own pages — chrome://, the Web Store, and the new-tab page are all off limits.
Fix: Open an ordinary website and try again. If it fails on a normal page, please report it with the address.
Why: Chrome hides new extensions behind the puzzle-piece menu by default.
Fix: Click the puzzle piece to the right of the address bar and pin Table Grabber. The keyboard shortcuts work whether it is pinned or not, and can be rebound at chrome://extensions/shortcuts.
| Keys | What it does |
|---|---|
| Alt+Shift+T | Open Table Grabber and list every table on the page |
| Alt+Shift+G | Grab a table by clicking it |
| Alt+Shift+B | Grab an area by dragging a box round it |
| ↑ / ↓ | While picking: widen or narrow the selection |
| Enter | While picking: take the current selection |
| Esc or right-click | While picking: cancel and leave the page alone |
| Double-click or F2 | In the editor: edit the selected cell |
| Enter / Tab / Esc | While editing: commit and drop down, commit and move right, or abandon |
| Alt+↑ / Alt+↓ | In the editor: move the selected row up or down |
| Alt+← / Alt+→ | In the editor: move the selected column left or right |
| Alt+Enter | In the editor: insert a blank row below |
| Delete | In the editor: empty the cell — Ctrl/Cmd+Delete deletes the row |
| Ctrl/Cmd+Z | In the editor: undo — Ctrl/Cmd+Shift+Z redoes |
| Ctrl/Cmd+S | In the editor: download in the selected format |
| Ctrl/Cmd+Shift+C | In the editor: copy to the clipboard |
| Ctrl/Cmd+F | In the editor: jump to the row filter |
A table that comes out wrong is genuinely useful to me — the awkward ones are how this got good. Please include:
Everything below is something Table Grabber answers directly.
Chrome Web Store extension — see every byte a site stores
Free SQL translator, formatter & validator
Git cheat sheet & fix-it playbook