On this page
Getting started
PullRows runs in desktop Chrome 116 or newer, Microsoft Edge and Brave. There's no account to create.
- Install. Open PullRows in the Chrome Web Store and click Add to Chrome. On Edge, use Edge Add-ons. Brave installs straight from the Chrome Web Store.
- Pin it. Click the puzzle-piece icon in the toolbar and pin PullRows so it's one click away.
- Open the side panel. Click the PullRows icon. The panel opens on the right and stays open while you browse, so you can scrape page after page.
Your first scrape in 30 seconds
- Open a page with a list — search results, a directory, a product grid.
- In the List tab, press Auto-detect. PullRows outlines the list it found.
- Check the preview columns, then press Start.
- Rows land in the Data Table. Press Export CSV.
Already had the page open before installing? Reload it once so PullRows can read it.
Tools
All eleven tools live in tabs along the top of the side panel and write into the same Data Table.
List Extractor
Turns any repeating list, grid or table into rows.
- Auto-detect scores every repeating block on the page and picks the most likely list. Not the one you meant? Cycle through the other candidates.
- Pick an item lets you click one item on the page; PullRows finds all its siblings. Press Esc to cancel picking.
- Columns are generated and named from the content — title, price, rating, reviews, date, link — even on sites with scrambled class names. Rename, remove or add columns by clicking elements.
- Pagination: none, Next button, Load more button or infinite scroll — including scrolling panels such as the Google Maps results feed. Set a page limit.
- Presets apply automatically on Google Maps, LinkedIn Jobs, Instagram (posts and Reels), YouTube and Amazon.
The scroll engine moves one screen at a time so lazy-loaded and virtualised items are captured, and merges rows that load in stages.
Page Extractor
Extracts the same fields from many URLs — one row per page. Paste URLs, or send them from the Sitemap Explorer or a Data Table column.
- Define fields by clicking elements on a sample page, or add built-ins: title, H1, description, canonical, Open Graph image, author, published date, language, word count, emails, phones, social links, price and main text.
- Presets add whole field sets: page metadata, contacts, product price, Google Maps place (name, rating, reviews, category, address, phone, website, hours) and Instagram post or Reel stats.
- Fast mode fetches each page's HTML in the background with your browser's cookies — quick, and logged-in pages work.
- Browser mode renders each page in a background tab. Use it for JavaScript-heavy sites and Google Maps.
Email Finder
Finds email addresses across a list of websites.
- Deep scan also visits contact, about, team and imprint pages.
- Decodes Cloudflare-protected addresses and spelled-out ones like
name (at) domain (dot) com, and filters out junk such as image file names. - Optional domain filter (only emails on the site's own domain) and phone number extraction.
Social Links
Finds LinkedIn, X, Instagram, Facebook, YouTube, TikTok, GitHub and other profiles for a list of sites. You get one row per site with a column per platform. Share buttons are ignored, so you get the company's own profiles.
Image Downloader
Collects every image on a page: srcset originals, lazy-loaded images, CSS backgrounds and the Open Graph image. Filter by size and type, pick what you want in a thumbnail grid, and download them into Downloads/pullrows/, in a folder per site, with readable file names. You can also download every image from a URL column in the Data Table.
Shopify Exporter
Exports a Shopify store's catalogue — all products or one collection — from the store's public product feed. Choose the format:
- One row per variant with price, compare-at price, SKU and availability.
- One row per product.
- Shopify import CSV, ready to import into another store.
Sitemap Explorer
Finds a site's sitemaps from robots.txt and common paths, follows sitemap indexes and .gz files, and groups URLs by section. Filter by text or a /regex/, then send the URLs to Page Extractor, Email Finder, Social Links or Page → Markdown.
Page → Markdown
Extracts the main content of a page — like reader mode — as clean text and Markdown, with title, author, date, language and word count. Works on the current page or a list of URLs. Copy the result or save it as a .md file.
Quick Scan
One click on the current page pulls emails, phones, links, images, tables, text, social profiles, prices and pagination links, plus anything matching a custom CSS selector. It can follow Next links across pages. Each type gets its own sheet.
AI features
Three optional features use Claude through Anthropic's API with your own API key (paste it in Settings and pick a model). Anthropic bills usage to your key.
- AI Extract — describe the table you want in plain English. Claude reads the page, or a list of URLs, and returns rows.
- AI column — in the Data Table, fill a new column from every row: “classify the industry”, “summarise in 10 words”, “extract the city”.
- Ask Claude — ask questions about your data and get an answer with numbers from the table.
Page text or table rows are sent to Anthropic only when you run one of these features. The Claude connector is separate and doesn't need an API key.
Data Table & exports
Every tool writes into the Data Table, a small spreadsheet inside the side panel.
- Sheets — each run gets its own tab; search across rows.
- Edit — double-click a cell to edit it.
- Column menu — rename, sort, move, dedupe by column, delete, send URLs to another tool, or download the column's images.
- Dedupe, Clean (trim spaces, drop empty rows and mostly-blank columns), Merge columns and delete rows.
| Export | What you get | Plan |
|---|---|---|
| CSV | UTF-8 CSV that opens cleanly in Excel, Numbers and Google Sheets | Free |
| Clipboard | Tab-separated, paste straight into any spreadsheet | Free |
| Excel (.xlsx) | One worksheet per sheet tab | Pro |
| JSON | An array of row objects | Pro |
Recipes & History
Save any tool's setup as a recipe — optionally with its start page — and re-run it in one click. Export recipes as JSON to share them or move them to another browser, and import recipes others share with you. The free plan keeps 3 recipes; Pro is unlimited. Claude can list and run your recipes through the connector.
History keeps your last 30 results on your device. Open one to load it back into the Data Table, or clear history at any time.
Monitors (change & price-drop alerts)
Save a recipe (for example List Extractor on a product, job or listing page, with a start page), then click 🔔 Monitor in Recipes. PullRows re-runs it every hour, day or week in a background tab while Chrome is open, compares it with the previous run and shows a Chrome notification for new rows, rows that are gone, changed values or only price drops — your choice. Each run is saved in History with a Changes sheet (change, item, column, before, after, percent). You can also ask Claude: “watch this page and tell me when anything gets cheaper”.
Google Sheets
In Settings → Google Sheets, click Copy setup script, paste it into your sheet's Extensions → Apps Script, deploy it as a Web app (Execute as: Me, access: Anyone) and paste the /exec URL back. Then Google Sheet in the Data Table appends the visible rows to a tab named after the dataset. With Pro, monitors can push their changes automatically. Cells that start with = are stored as text, so scraped content can't run formulas.
AI enrichment
✦ Enrich in the Data Table (Pro, with your Anthropic key) adds several columns at once: lead score (0–100 against your ideal customer) with the reason, industry, company size and B2B/B2C; or summary and category; sentiment and topic; first/last name and role; or a translation. Unknown values are left empty instead of guessed.
Claude connector
Background by default. Claude’s tools read a page’s HTML with no tab at all when the data is in it (and follow “Next” links); a script-built page is opened in a background tab you are never switched to and closed afterwards. Each result includes how_it_ran, so Claude can tell you what it read and where. Ask Claude to use mode “browser” when you want to watch the page and get a screenshot in the chat.
The connector lets Claude use PullRows' tools through MCP, the Model Context Protocol. Claude asks; your browser does the work, with your logged-in sessions; results come back to Claude and also land in your Data Table.
Claude ──MCP (HTTPS)──▶ PullRows relay ◀──── PullRows in your Chrome
passes calls through, does the scraping,
stores nothing returns results
Claude's servers can't reach your browser directly, so a small relay sits in between. It holds tool calls and results in memory only long enough to pass them on. PullRows uses our hosted relay by default; developers can run their own and enter its address under Relay server in Settings.
Turn it on
- Open the side panel → Settings → Claude connector.
- Switch on Allow Claude to use this browser. The status changes to ● Connected.
- Press Copy next to your private link. It looks like
https://…/mcp/<secret>.
claude.ai and Claude Desktop
- In Claude, open Settings → Connectors and choose Add custom connector.
- Name it PullRows, paste your private link and press Add.
- In a chat, enable PullRows from the tools menu, then ask away.
Connectors you add on claude.ai also appear in Claude Desktop when you're signed in to the same account. Custom connectors need a Claude plan that supports them.
Claude Code
claude mcp add --transport http pullrows <your-link>
Run /mcp inside Claude Code to check the connection.
Tools Claude gets
| Tools | What they do |
|---|---|
browser_tabs, browser_open, browser_screenshot, page_action | See and drive pages — list tabs, open a URL, screenshot, click, scroll, type |
page_read | Page as clean Markdown plus metadata |
list_detect, list_extract | Lists, grids and tables with Next, Load more or infinite scroll |
page_extract | The same fields from many URLs, with built-ins like Google Maps place fields |
find_emails, find_social_links | Contact crawling over a list of sites |
shopify_products, sitemap_urls, images_find, images_download, quick_scan | The other side-panel tools |
data_table_read, recipes_list, recipe_run | Read results and run your saved recipes |
Prompts to try
- “Open this Google Maps search and give me every cafe with its rating and phone number.”
- “Find the contact emails for these 20 company websites.”
- “Export this Shopify store's products and tell me the median price per collection.”
- “Run my competitor prices recipe and compare it with last week's numbers.”
Each tool call has about four minutes to finish. For big jobs, ask Claude to work in chunks — say, 5 pages at a time. On the free plan the connector allows 25 tool calls a day.
Safety controls
Letting an AI use your browser should be boring and predictable. These are the guardrails:
- Off by default. Nothing connects until you switch the connector on. Switching it off disconnects immediately.
- A private link. The secret in your link is the only key to your browser. Don't share it. Generate new link revokes the old one instantly.
- Read-only by default. Claude can open pages, read them and extract data. Clicking, typing and file downloads stay blocked until you allow Actions and Downloads in Settings → Claude connector.
- Sensitive sites are blocked. Claude can never open, read or screenshot webmail and account pages (Gmail, Outlook, Google, Microsoft and Apple accounts), payment and banking sites (PayPal, Stripe, Wise, Revolut and any site with “bank” or “wallet” in its address), password managers (1Password, LastPass, Bitwarden, Dashlane), the Chrome Web Store, browser pages and
localhostor private-network addresses. Edit the list under Blocked sites in Settings. These rules apply only to Claude — your own scrapes are unaffected. - You confirm before it starts. Turning the connector on shows exactly what will pass through the relay, and asks for your OK.
- Everything is visible. Settings shows Claude's recent activity, and results Claude collects also land in your Data Table, so you can see exactly what was read.
The relay passes calls and results through in memory and doesn't store page content. See the privacy policy for details.
Plans & limits
| Free | Pro & Lifetime | |
|---|---|---|
| All 11 tools, Data Table, presets | Yes | Yes |
| CSV & clipboard export | Unlimited | Unlimited |
| Excel & JSON export | — | Yes |
| List pagination | 10 pages per run | No limit |
| Page Extractor | 25 URLs per run | No limit |
| Email & Social finder | 25 sites per run | No limit |
| Images / Shopify / Sitemap | 100 / 100 / 1,000 per run | No limit |
| Recipes | 3 | Unlimited |
| AI features | With your Anthropic key | With your Anthropic key |
| Claude connector | 25 tool calls a day | Unlimited |
| Monitors | 1, daily | Unlimited, hourly |
| Google Sheets | Send rows | Send rows + auto-sync monitor changes |
| AI lead enrichment | — | Yes (your Anthropic key) |
| Browsers per licence | — | Up to 3 |
After checkout you'll get a licence key by email. Paste it in Settings → Plan. Each browser activates once and re-checks the licence about once a day; if the licence server can't be reached, Pro keeps working for 14 days. Payments are handled by Dodo Payments, our merchant of record. Every purchase has a 14-day money-back guarantee.
Troubleshooting
A list stops early or items are missing
Chrome pauses lazy loading in hidden windows and background tabs. Keep the tab you're scraping visible — not minimised or covered — while a scroll job runs. If you're working with Claude, put Chrome side by side with it. PullRows warns you when the tab was hidden during a run.
Page Extractor returns empty fields
The site probably builds its content with JavaScript, or blocks background requests. Switch the run to Browser mode, which renders each page in a real tab.
“Can't run on this page”
Chrome doesn't let extensions read its own pages (chrome://), the Chrome Web Store or some built-in viewers. On ordinary sites, reload the page once after installing or updating PullRows.
Claude says the connector is offline
Chrome needs to be open with the connector switched on — check that Settings shows ● Connected. If you generated a new link, update the connector in Claude with the new one.
A Claude tool call timed out
Ask for smaller chunks, such as fewer pages per call or 20 URLs at a time.
My licence key isn't accepted
Check for extra spaces when pasting. A key works on up to 3 browsers; if you've reached that, email support@canwe.tech and we'll reset it.
Keyboard & accessibility
- Shortcut: set a keyboard shortcut to open PullRows at
chrome://extensions/shortcuts. - Navigation: the side panel uses standard buttons, inputs and menus, so Tab, Shift+Tab, Enter and Space work throughout, with visible focus.
- Editing cells: Enter saves, Shift+Enter adds a new line, Esc cancels.
- Picking elements: Esc cancels the element picker.
- Reduced motion: this website respects your system’s reduced-motion setting.
Found something hard to use? Tell us at support@canwe.tech.