Onward

Onward is a small auto-pager userscript. When you reach the bottom of a paginated page (search results, forum threads, blog archives, shop listings) it finds the link to the next page, fetches it, and adds that page's results below the ones you're reading. You keep scrolling instead of clicking "Next".

It does the same job as Pagetual and AutoPagerize, but the whole thing is about 1,400 readable lines instead of 13,000. There's no rule database to download. Detection works from the page itself, and you can teach it a site in two clicks when it guesses wrong.

1. Install Tampermonkey or Violentmonkey.
2. On Chrome, Edge and other Chromium browsers, let the manager run scripts. Managers built on Chrome's user scripts API, Tampermonkey 5.5 and later among them, run nothing until you do, and nothing tells you why.
   • If the manager's Details page (from chrome://extensions or edge://extensions) has an Allow User Scripts switch, as Chrome 138 and later do, turn it on.
   • Otherwise turn on Developer mode at the top of the extensions page.
3. Open src/onward.user.js and accept the install prompt.

The script updates itself from this repo. Firefox needs no extra step.

Userscripts for Safari should work too: Onward keeps its settings through the async GM.* storage that manager offers, and since it has no menu, a small "Onward" button in the corner (and an "Onward…" button on each page bar) opens Settings, where the menu's commands are. That hasn't been tried on a real iPhone or Mac yet.

• Finds the next page without rules. It checks <link rel="next"> first, then scores links and buttons by their label ("Next", "Older posts", 下一页, 次へ, 다음, Следующая and about 80 more), their classes, whether they sit in a pager, and whether the URL is the current ?page=N or /page/N/ plus one. Previous-page links, disabled items, carousels and sliders are ruled out. Numbered pagers work too: if page 3 is marked current, the link labelled 4 wins.
• Finds the list to extend. It looks for the element whose children repeat (results, posts, rows, cards) and covers the most of the screen, skipping headers, footers, menus and the pager itself. Interleaved rows like Hacker News' title and subtext lines stay together.
• Adds only new content. Items that repeat word for word from page 1 (sticky threads, "sort by" bars, headings) are dropped. If a site sends back a page it already sent, Onward stops.
• Handles the awkward cases. Lists rendered by JavaScript get loaded in a hidden iframe. "Load more" buttons get clicked. Pages that scroll inside a div instead of the window work. Non-UTF-8 pages (Shift_JIS, GBK, windows-1251) are decoded properly, and lazy-loaded images get their real src.
• Survives React and friends. If the site redraws its list and throws away the added pages, Onward switches to putting each page in its own copy of the list, which frameworks leave alone.
• Stays out of the way on sites that already scroll forever. Discourse and Flarum forums are left alone. On other sites Onward watches for a few seconds before its first page and whenever you reach the bottom. If the site adds items by itself, Onward takes its own pages back out and stands by.
• Is polite to servers. At most 40 pages per visit by default. A failed request pauses loading until you click Retry. If added pages don't make the page any longer, it stops rather than loading forever.

Each added page gets a slim bar with the page number, its URL, and buttons to jump to the top, to stop, or to skip to the footer (loading then waits 30 seconds, so you can reach it). Once stopped, the bars offer Resume instead. As you scroll, the address bar follows the page you're reading, so refreshing or sharing the link lands in the right place.

Open your userscript manager's menu on any page:

When automatic detection gets a site wrong, the picker is the quickest fix. You can also write rules by hand in Settings. A rule is a JSON object:

[
  {
    "url": "^https://forum\\.example\\.com/threads/",
    "next": "a.pagination-next",
    "content": "#posts > .post",
    "insert": "",
    "mode": "fetch",
    "click": false
  }
]

• url is a regular expression matched against the page address.
• next and content accept CSS selectors or XPath (anything starting with /, ./, ( or id().
• insert is optional. New pages go before the first element it matches.
• mode is auto, fetch or iframe. Use iframe for sites that build their lists with scripts.
• Set click to true when next is a button with no link, such as "Load more".
• excludeUrl is optional. The rule is skipped on addresses it matches.

A rule only takes over on pages where its selectors actually find something. On other pages of the same site, Onward falls back to finding the next page by itself.

Rule lists in AutoPagerize/wedata format (url, nextLink, pageElement, insertBefore) are understood too. Paste their URLs into "Rule list URLs" and Onward refreshes them weekly. Lists on wedata.net, hoothin.github.io and cdn.jsdelivr.net load straight away. For a list anywhere else, Tampermonkey asks you once whether Onward may connect to that site. Very broad patterns from those lists are skipped so they don't override detection.

At the bottom of the panel, Diagnostics shows what Onward found on the page you're on: how it found the next link and where that leads, the list and how many items it started with, the loading mode, the rule it used and where that rule came from, and the last error. "Copy diagnostics" puts all of it on the clipboard for a bug report. It includes the page's address and the next one, which can hold private details, so read it over before you paste it. Nothing is sent anywhere.

Other scripts on the page can follow what Onward adds. It sends the same events as AutoPagerize, so scripts written for AutoPagerize or weAutoPagerize keep working.

Every added element carries the class autopagerize_page_element (in wrapped mode, the page's copy of the list does).

Onward listens on document for AutoPagerizeDisableRequest, AutoPagerizeEnableRequest and AutoPagerizeToggleRequest, so another script can pause paging and start it again:

window.addEventListener('onward:page', (e) => console.log('page', e.detail.page, 'from', e.detail.url));
document.dispatchEvent(new Event('AutoPagerizeDisableRequest'));

The tests need Node 22.22.2 or later on the 22 line, 24.15 or later on the 24 line, or 26 and up (what jsdom 30 supports).

npm install
npm test          # detection unit tests (jsdom)
npm run e2e       # the real script in headless Chromium against a local fixture site
npm run e2e:isolated   # the same, with the script in an isolated world as userscript managers run it
npm run check     # syntax check and version consistency
npm run screenshots
npm run smoke     # real sites from scripts/smoke-sites.json, checked against what Onward should do there
npm run smoke -- https://news.ycombinator.com/news   # or just report on any page
npm run bench -- items_all.json   # what a wedata-sized rule list costs (download it from wedata.net first)

npm run e2e needs Playwright's Chromium (npx playwright install chromium if it isn't there). npm run smoke loads at most three pages per site and fails only when Onward does something other than expected. A site that shows the headless browser a bot check, or refuses its requests, is reported as not checked, and Google and Bing are on the list as things to check by hand. The fixture site in test/e2e/site.js covers a paginated blog, a windows-1252 forum table, a script-rendered grid, a list that redraws itself, a failing server, an inner scroll container and a load-more button.

The script is one file on purpose. It's wrapped in a small factory so the same code runs in the userscript manager and in Node tests without a build step.

MIT. See LICENSE.