Create your first PromptLoop task

Q: Can I edit a task after creating it?

Yes. Click Edit on any task to modify queries, formats, or crawl depth. Changes apply to new jobs only.

Q: How many output columns should I include?

Start with 3-5 output columns. More than 6 columns can slow down the task or reduce accuracy.

A PromptLoop task is a no‑code AI agent that extracts data from websites. Give it a list of URLs or company names and it returns clean, structured columns—at any scale.

Prerequisites

A PromptLoop account (creating and editing tasks is free on all plans)
A sample URL or company name to test with

Choose your task type

Task Type	When to Use	Example
Website	You have a list of URLs	Extract pricing from competitor websites
Search	You only have a name or keyword	Find LinkedIn URL for "Acme Corp"

Steps

1. Start with a template (fastest way)

Go to Tasks → New Task or visit the template library
Browse templates and click one that matches your goal (e.g., "Find Information from Website")
Click Use Template to create a copy you can customize

Create task from template or AI generation

2. Or let AI generate it for you

Click New Task → Generate with AI
Describe what you need: "Extract company headcount and industry from websites"
Review the auto‑generated output columns and click Save

PromptLoop uses your description to create the task structure—saving you setup time.

3. Test your task

Open your new task and paste a sample URL or name in the test input
Click Run Test
Review the results and hover over the info icon to see the AI's reasoning

4. Customize output columns (optional)

Click Edit to modify what data you extract:

Name: Column header (keep it short and unique)
Query: What you want to find (e.g., "What is the main phone number?")
Format: Choose Text, Number, True/False, Category, List, or Link
Instructions: Add rules like "Keep under 10 words" or "Return 'Unknown' if not found"

Pro tip

Start with 3-5 output columns. Tasks with too many outputs can run slower or lose accuracy.

5. Set crawl depth

Choose how deep the AI should explore each website:

Single Page: Only the exact URL you provide
Smart Crawl (recommended): Follows a few relevant links automatically
Deep: Explores many pages (uses more credits)

Most tasks work best with Smart Crawl.

Run it on a dataset

Once your task works on test inputs:

Upload or select a dataset with your URLs or company names
Click Launch Job and pick your task
Map your columns to the task inputs
PromptLoop adds new columns with the extracted data

View and export results from the Datasets tab.

Quality checks

Task returns results for your test inputs
Output format matches what you need (text, numbers, categories, etc.)
Column names are clear and unique
Results include sources when you need to verify data

FAQs

Can I edit a task after creating it?
Yes. Click Edit on any task to modify queries, formats, or crawl depth. Changes apply to new jobs only.

How many output columns should I include?
Start with 3-5. More than 6 columns can slow down the task or reduce accuracy.

What if my task returns blanks?
Test with a single-page crawl first to debug. The data might be on a different page, or you may need to adjust your query.

Can I reuse tasks across different datasets?
Absolutely. Once a task is saved, you can run it on any dataset with matching input columns.