Schedule a web scrape with no code and get only the new rows

What is a scrape task, and how is it different from a rule?

A scrape task is a job that opens a page on a schedule, picks the values you marked on it (a name, a price, a link, a date), follows the next pages, keeps only the rows it has not sent before and delivers them where you said. Minda Super Intelligence runs it on your computer, in a hidden page of the workspace you made it in, so it reads the page as you see it, sign-in included, and it needs no code and no AI model: a plain scrape costs no Minda AI credits on any plan.

A rule (an automation) is the other tool: it watches one part of a page, a schedule, your computer or your messages and acts when something holds. A rule answers "tell me when this changes"; a scrape task answers "give me every new row of this list". If what you want is one price or one status word, read how to monitor a web page for changes instead. A cloud automation is a third thing: it runs on Minda’s server while your computer is off, but only on public pages and APIs, and it never reads a page you are signed in to.

What can a scrape task collect, and from where?

• The page’s main list, found for you. Extract data finds the list on the page (a table, a grid of cards, search results) and its columns; keep the ones you want and rename them.
• Your own fields, picked with the pointer. Press Pick, point at the first title, the first price, the first link: the click becomes a selector that matches the same value in every item ("Found in 10 of 10 items"), with no selector to type. A field can be text, a link, an image or any attribute, and a CSS selector of your own works too.
• Next pages. The task follows the page’s Next link, or a selector you give, up to 20 pages, a second apart, on the page’s own site only; a row already seen on an earlier page is dropped.
• Only new rows. Tick Only new rows and each run sends the rows no run had before ("28 rows, 1 new"), so a daily run of a job board or a listings page gives you the new entries, not the whole list again. A quiet run announces nothing unless you ask for the rows after every run.
• Pages behind your sign-in. The run uses the workspace’s own sign-ins, so a supplier portal, a client dashboard, an orders page or a members-only list is read the same way as a public page. Pages you are signed in to are never read in the cloud.

How do I set up a scheduled scrape?

1. Open the page in Minda, in the workspace whose sign-in it needs, and go to the first page of the list you want.
2. In the Tasks panel choose Scrape a page on a schedule (No AI), or start from a template: Collect leads every day (Name, Phone, Address, Email, Website; 5 pages; only new rows; daily at 09:00; a note and Excel) or Track a price every hour (Name, Price, Change; hourly; every run; your API). Extract data has a Schedule button for a list you have already chosen.
3. Keep the list Extract data found, or press Pick and point at your own fields; name each one.
4. Set the next pages: the task follows Next on its own, or you point at the link, and you choose how many pages.
5. Choose Send results to: a note, an Excel or CSV file (Word or PDF for a document), your own API, email or WhatsApp.
6. Set the repeat: every 15 or 30 minutes, every hour, every 2 to 12 hours, every day, every weekday, a cron line of your own, or once at a time you pick. The editor shows the next three runs.
7. Save and press Run now. The first run reports what it found; from then on, each run sends only what changed or is new.

Where can the rows go?

Five destinations, each your own, set on the task and never chosen by a page or a model. Each run sends only new or changed rows unless you tick Send the rows after every run.

How often can it run, and what happens while it runs?

A schedule runs from every 15 minutes to a cron line of your own, in your time zone, while Minda is open on that computer; it can also run once at a time you pick, which is what "in 10 minutes, check this page and tell me" in the AI chat makes. A run needs no window open to the page: it loads the page in a hidden page of the workspace, with that workspace’s sign-ins, and a time missed while Minda was closed runs once after launch. Each run stops after a minute a page, keeps at most 5,000 rows and 50 columns, and stays on the page’s own site: a next-page link elsewhere is refused.

Two guards hold for every scheduled run. Ask before paying is on by default, so a run that reaches anything that pays, buys or places an order stops for your OK (turned off, a run pays too), and a run that waits tells you. And a scheduled run never clicks or types on chat, social or mail sites, whose terms ban automated sending: it can message you with the rows, not your customers.

Timers and reminders are separate and free: "remind me at 15:30 to call Bob" is a notification with a sound, no AI run, no credits, and never counted as a scheduled task.

Does a scheduled scrape cost anything?

No. Scheduled tasks and scrape jobs are unlimited on every plan, and a plain scrape, which reads values and sends rows, uses no AI model and no Minda AI credits. The Free plan runs as many as Pro.

Credits enter only when you add an SI step: a daily brief that reads the new rows and writes a summary, a meeting prep built from your calendar, a weekly review. Those are the AI templates of scheduled tasks, and each AI turn in them is a request on Minda’s hosted models (100 credits a month on Free, 2,000 on Pro) or free with your own OpenAI, Anthropic or Gemini key or Ollama on your computer. Keep the scrape plain and it stays free; add a summary and you pay by use.

Scrape task, rule or cloud automation: which one do I need?

Pick by what you want from the page and where it has to run.