Dataset steps, fields and transforms
A dataset’s steps are what each run does on each page, in order, and they are the same on every run. This page lists every step in the editor’s Add step menu, how a step finds its element, and the fields, transforms and row key of the steps that collect data.
Datasets are in early access. If your account doesn’t have them yet, open Datasets in the app and click Ask for early access, or write to us through the contact page.
Add and arrange steps
- Click Add step under the steps. A menu opens with every step type by group: Navigate, Interact, Wait, Data and Advanced. Type to search it.
- The new step opens. Fill in its options, give it a Step name if you like, and click Done.
- Drag a step by its handle to move it. The ⋯ menu on a step has Run to here, Add a step below, Duplicate, Turn off (runs skip it until you turn it on), Move up, Move down and Delete.
A dataset has up to 50 steps, and up to 20 inside For each item.
Step types
Navigate
| Step | Does |
|---|---|
| Go to | Opens an address, a path on the input’s site (such as /blog/page/2), or the input itself with Use the input ({{input.url}}), so one set of steps covers every address in Inputs. |
| Back | The browser’s back button. |
| Reload | Loads the same page again. |
Interact
| Step | Does |
|---|---|
| Click | Clicks a button, a link or a cookie banner, with One click, Double-click or Right-click. |
| Type | Fills a field, like a search box, with Text to type. It can Clear the field first and Press Enter after. The text can use an input’s values, such as {{input.term}}. |
| Press a key | Enter, Escape, Tab, ArrowDown, or a shortcut like Control+A. |
| Select | Chooses an option from a dropdown, by its text or its value. |
| Check, Uncheck | Ticks or clears a checkbox or a radio button. |
| Hover | Opens a menu that shows on hover. |
| Scroll | To the bottom of the page, to the top, to an element, or by a number of pixels. |
Wait
| Step | Does |
|---|---|
| Wait | A fixed pause, in seconds. |
| Wait for | Goes on as soon as an element shows (or goes), the address changes, or the page stops loading. Steadier than a fixed pause. |
Data
| Step | Does |
|---|---|
| Collect list | One row for each item of a list: cards, results, table rows. Each item is the element that repeats; its fields are read inside each item. |
| Collect one | One row from the page, such as a product or an article. Inside For each item, its fields join the item’s row. |
| Next page | Reads the pages after the first and repeats the nearest Collect list on each (see below). |
| For each item | Opens each item of a Collect list by its link and runs the steps inside it on the item’s page (see below). |
| AI read | Reads the page with AI from a description in words (What to read), with optional field names. It gives Rows, or Values for this row. It costs 5 credits more for each page it reads; for pages with structure, Collect list is exact and costs nothing extra. |
Advanced
| Step | Does |
|---|---|
| Run code | JavaScript that runs inside the page with a time limit, no network calls and no files, and returns JSON: Rows: a list of objects, or Values for this row. It costs nothing extra. |
How a step finds its element
A step that acts on an element, such as Click, Type or Collect list, has a target. Click Pick on page and then the element on the preview: Monity writes the target for you. To write it yourself, choose how to find it: CSS, Text, Role, Label, Placeholder, Test id or XPath.
Monity keeps other ways to find the same element. When a page changes and the target finds nothing, a run tries those, and the run’s record says Worked, after finding its element another way.
Under When it goes wrong, each step says what a run does if the step doesn’t work: Stop the run (the default), Go on without it (for what shows only sometimes, like a cookie banner) or Go on, with a warning. Wait for it up to sets how long the step waits for its element.
Fields
Collect list and Collect one have a fields table, one field a column of your rows. A step has up to 50 fields.
| Column | Means |
|---|---|
| Name | The column’s name in the rows, alerts and exports. |
| Take | What to read: Text, Link, Image, Attribute (such as href or data-id), HTML or Markdown. |
| From | The element inside the item, picked on the page. |
| Transform | How the value is cleaned, applied in order on every run (below). |
| Sample | The value read on the preview, after the transforms. |
| Key | Marks the field that names an item, usually its link (see Row key). |
On Collect list, Suggest fields adds the fields Monity sees in the item that are not there yet, and One level up or One level down changes which element is the item. Details on a field holds its target and transforms, and whether to skip an item that has no value for it (in a list) or fail the step when the value is missing (on one page).
Transforms
| Transform | Does |
|---|---|
| Trim | Squeezes spaces and line breaks. |
| Date → ISO | Writes a date as 2026-10-20. Format on the page says how the page writes it, such as MMM D for “OCT 20”, or auto. Year missing: Use this year or Leave the field empty. |
| Number | “12,500” becomes 12500. Decimal mark: Work it out, Point: 1,299.50 or Comma: 1.299,50. |
| Price | “$1,299.00” becomes 1299: the currency sign is dropped. |
| Pattern | Keeps the part that matches a regular expression, such as (\d+) min, and the Group you choose. No match leaves the field empty. |
| Absolute address | “/blog/a” becomes https://…/blog/a. |
A field has up to 10 transforms.
Row key
The fields marked Key (up to 5) name a row. The same key on the next run is the same row: a new value in another field makes it Changed, not New, and a key that is no longer there is Gone. Monity ticks Key on the item’s link when it proposes the fields. With no key, a row only matches a row that is the same in every field.
Next page
Reach the next page by:
- Next button: clicks the Next button or link, then repeats.
- Page numbers in the address: opens
…/page/2,…/page/3and so on. The page’s address has{page}where the number goes, and The second page is number says where it starts. - Load more button: clicks Load more; new items join the page.
- Scrolling down: scrolls to the bottom; new items load as you go.
Read up to sets how many pages, the first included (up to 100, and never more than the Pages limit on Settings). Each page repeats from chooses the step to repeat, the nearest Collect list by default. A run stops at the last page, at that number, or at a page with nothing new. A dataset has one Next page step.
For each item
Open the items of chooses the Collect list, and By its link in the field with each item’s address. The steps inside it, usually a Collect one, run on each item’s page, and their fields join the item’s row. Add them with Add a step inside.
A run opens one page for each row, on top of the list’s pages, at 1 credit a page. When the dataset has a row key, an item opened in the last 7 days keeps its values and isn’t opened again, as long as the steps inside haven’t changed; after 7 days it is opened again. Run to here stays on the list.