Skip to main content
The Scrape page in your dashboard lets you run a scrape from a form. Give it up to 3 URLs, say what you want, and read the result on the right. Anything you do here can also be done with the API, an SDK or an AI assistant through the MCP server. This page walks through one example from start to finish: the title and price of every book on books.toscrape.com, a practice site built for scraping. The Scrape playground with the form on the left and the Output panel on the right

​
Run your first scrape

  1. Open Scrape in the sidebar.
  2. Enter https://books.toscrape.com in Target URL and press Enter.
  3. Write Return the title and price of every book on the page in Extraction instruction.
  4. Choose JSON under Output.
  5. Click Start Scraping.
About ten seconds later the books appear in the Output panel. The Output panel showing a JSON array of books, each with a title and a price This is the real result, shortened. The page has 20 books:
The run used 3 credits: 1 for the page and 2 for the AI extraction, from about 3,000 tokens. Switch the panel from JSON to Table to see the same data as rows, and use the copy and download buttons to take it with you.

​
The form, control by control

The Scrape form with numbers marking each control from 1 to 9

​
1. Target URL

Type or paste a URL and press Enter. It becomes a chip. Add up to 3 URLs and Spidra combines what it reads from all of them into one result, which is useful for comparing pages or answering one question from several sources. If you want a separate result for each URL, use Batch scrape.

​
2. Operations

Each URL has an Add operation button. Operations are steps Spidra performs on the page before it reads it: click a button, type into a field, scroll, wait, or loop over every item with forEach. They suit pages that hide content behind a cookie banner, a Load more button or a search box. A URL with operations is marked AI or CSS. In AI mode you describe the element in plain English, such as Click the accept cookies button. In CSS mode you give a selector, such as #submit-button, when you want exact control. See Browser actions for every operation type and worked examples.

​
3. Extraction instruction

This is the prompt. Describe the data you want in plain language and name the fields.
  • Weak: Get the books
  • Better: Return the title and price of every book on the page
Leave it empty and Spidra skips the AI step and returns the page as text, which costs 1 credit and uses no tokens.

​
4. Output

​
5. Schema

Define schema opens an editor where you describe the exact shape of the result. Use it when the result feeds a database or another program and the field names and types must not change. The JSON Schema editor with title, price and in_stock fields Run against a single book page, a schema with title, price, currency, in_stock and rating returns exactly those fields:
price is now a number and not the string "£51.77". rating is null because that site stores the star rating in the page’s HTML and not in its text. This run cost 2 credits. Read Structured output for how to write schemas.

​
6. Stealth Mode

Sends the request through a residential proxy, and you can choose the country. Turn it on for sites that block ordinary traffic or show different content by region. It uses proxy bandwidth, which is separate from credits. See Stealth Mode.

​
7. Extract content only

On by default. It removes menus, headers and footers so the AI reads the main content of the page. Turn it off if the data you need lives in the navigation or footer.

​
8. Authentication

Configure lets you pass your own session cookies so Spidra can read pages that sit behind a login. See Authenticated scraping.

​
9. Fast Mode

A switch at the top right. With Fast Mode on, Spidra fetches the page over plain HTTP instead of opening a browser. It is quicker for static pages, and it turns off operations, screenshots and Stealth Mode. If a page needs a browser, Spidra falls back to one automatically.

​
Get the code for any run

Click Code at the top right and Spidra generates the matching request for the settings in your form, in JavaScript, Python or curl, using the API key you select. The Code Snippets drawer showing a JavaScript fetch call for the books.toscrape.com scrape Run it, and the response gives you a jobId. Fetch the job to read the result, as shown in the Quickstart.

​
Reuse a setup

Save as Preset stores the URL, instruction, schema and options so you can rerun the same scrape with one click. See Presets, or find ready-made ones in the Marketplace.

​
When the result isn’t what you expected

  • The output is empty or {}. The page may not contain what you described, or it loads content after the first view. Add an operation to scroll or click, or check the page in the Screenshot output.
  • A field is null. The value isn’t in the page’s text. Find it in the page first, then adjust the instruction.
  • The scrape fails. You aren’t charged for a failed job. Check Logs for the error.
  • The page is blocked. Turn on Stealth Mode.
Every run is saved in Logs with its output and the credits it used.

Crawl a site

Follow links and get one result per page

How credits work

What each scrape costs