> ## Documentation Index
> Fetch the complete documentation index at: https://parabola.io/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Pull from Browser

> Use Prowork's Pull from Browser step to open a cloud browser, follow plain-English instructions on any website, and return scraped data or a downloaded CSV.

<Note>
  This step is only available on certain plans. If you are interested in using this step, please reach out.
</Note>

The **Pull from Browser** step opens a secure cloud browser and follows your plain-English instructions on a website. It either pulls back the data it reads from the page or imports a CSV file it downloads. Use it for sites that don't have an API or a Prowork integration: export a report from a vendor portal, check prices on a supplier's site, or copy a table from a dashboard.

You'll find it in the step library under **Utilities**.

## Set up the step

<Steps>
  <Step>
    Add **Pull from Browser** to your flow.
  </Step>

  <Step>
    In **Starting URL**, enter the page the browser should open first.
  </Step>

  <Step>
    In **Prompt**, describe what the browser should do and which data you want back. Write it the way you'd brief a colleague.
  </Step>

  <Step>
    Under **Import**, choose one:

    * **Downloaded file** (default): imports a CSV file the browser downloads.
    * **Scraped data**: returns the data the browser reads off the page, as rows and columns.
  </Step>
</Steps>

<Tip>
  Example prompt: "Go to Reports, set the date range to last 7 days, and download the Orders CSV."
</Tip>

Leave **Keep the same columns on every run** checked. The step reuses the columns from its first run, so a run that finds nothing won't break the steps after it. Uncheck it if the columns you're pulling have changed.

## Sign in to a site

By default, the browser runs signed out. For a site that needs a login, contact Parabola support to set up a saved browser sign-in. Support will give you a **Kernel Profile ID**. Connect it to the step, and the browser starts already signed in.

If the site asks you to sign in again, the step stops and shows a **Re-authenticate** link. Click it, sign in, then run the step again.

## Run once per row

Connect a step with input data to run the browser once for each row. Reference any column in the URL or the prompt with `{Column Name}`:

* URL: `https://example.com/search?q={SKU}`
* Prompt: "Search for `{SKU}` and return its price."

When you run once per row, set **Import** to **Scraped data**.

Results include your input columns, then the scraped columns. If a row fails, an **Error** column explains what happened. A failed row doesn't stop the others.

## How it runs

The step opens a fresh browser on every run, whether you run it in the editor, on a schedule, or from a published flow. The browser goes to your starting URL, follows your prompt, then closes and passes the results to the next step. It knows today's date, so prompts like "last 7 days" work on scheduled runs.

With **Scraped data**, you get one row for each item the browser finds, with a column for each field. If the browser can't organize what it read into rows, the step returns a single `answer` column with its text reply.

## Limits

| Limit | Value |
| - | - |
| File downloads | CSV only. Available only when the step has no input connected. |
| Time per browser task | 10 minutes |
| Rows when running once per row | The first 100 input rows |
| Total time when running once per row | About 40 minutes. Rows that can't start in time are marked as skipped. |

## FAQ

<AccordionGroup>
  <Accordion title="The step returned no data. Why?">
    The browser may have hit a login page, a captcha, or a page that looks different from what your prompt describes. Make your prompt more specific about where to click and what to read. If the site needs a login, check that your sign-in is connected.
  </Accordion>

  <Accordion title="Can it get past captchas or two-factor codes?">
    Not on its own. For sites with two-factor sign-in, contact Parabola support about setting up a saved sign-in.
  </Accordion>

  <Accordion title="Can I watch the browser while it runs?">
    Not yet. You'll see the results once the step finishes.
  </Accordion>

  <Accordion title="Can it download Excel or PDF files?">
    Not yet. Downloads are CSV only. For other kinds of pages, use **Scraped data**.
  </Accordion>

  <Accordion title="Will a site block it?">
    It might. Some sites limit automated visits or allow only one session per account at a time. If a run fails partway through, try fewer rows or space out your scheduled runs.
  </Accordion>
</AccordionGroup>