v3.0 — Two-Click Lists & On-Device AI

Released: October 19, 2026

What's New

Pick a list in two clicks, on slow computers too

  • The picker reads the section under your cursor and its own cards, and is 2–3× faster on slow machines.
  • Fields are typed (prices, ratings, review counts, IDs) and split where a site packs several values together. They're found again by their meaning when a site changes its layout.
  • Stop always ends a run and keeps the rows read so far. Runs that follow a Next link continue across pages reliably.

Same details from many pages, out of sight

  • Many URLs run in hidden background tabs, with no windows over your work.
  • Steps run in your recipe's order, clicks are checked for an effect, and Job details shows coverage per field.
  • CrawlPilot suggests how many tabs at a time suit your computer, and eases off when it struggles.

Page assistant (on-device AI)

  • A brief made for the kind of page (product, article, listing, job, recipe…), with every point checked against the page and shown on it.
  • Answers with citations, as tables you can save, in your language.
  • Ask about an image or chart on the page.
  • What changed since I saved this? and Does it fit me?

It runs on Gemini Nano, Chrome's built-in model. Nothing leaves your computer. See Page assistant.

New tools

  • Contacts from websites: emails, phones and social profiles from a list of company sites, one row per company.
  • API recipes: the Scraper Toolkit finds the JSON API behind a page, and Get every page calls it page by page with your session, without a tab per page.
  • Live Google Sheets sync: send a dataset to your own Google Sheet through an Apps Script you own, with no CrawlPilot account. New rows follow automatically. See Settings.
  • Image Downloader: full-size originals, duplicates removed, convert and resize.
  • Markdown of any page, and AI Extract with grounded values.

Faster and lighter

The side panel opens about 40% faster. Only English is bundled; other languages load when you choose them.

Privacy

CrawlPilot collects nothing: no analytics and no account. Pages, extracted data, briefs and AI prompts stay on your device. Data goes only where you send it: your own Google Apps Script, webhooks you configure, and the sites you scrape (with your own session).

Upgrade Notes

  • Access to websites is asked for from a banner in the side panel: click Allow once. See Permissions.
  • The Text Summarizer is now the Page assistant. It no longer sends page text to an external API.
  • The Metadata Extractor is now Article text & details inside Same details from many pages.