Skip to main content
POST

Authorizations

Authorization
string
header
required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Body

application/json
urls
object[]
required

1 to 3 pages to scrape, each as an object such as { "url": "https://example.com" }. Plain strings aren't accepted here.

Required array length: 1 - 3 elements
prompt
string

What you want from the pages, in plain English, such as "Return the title and price of every book". Without a schema, the AI picks the shape of the answer.

schema
object

A JSON Schema for the answer. You get exactly these fields and types, with null for anything the page doesn't show. The root must be "type": "object", at most 5 levels deep and 10 KB. Sending a schema sets output to "json".

output
enum<string>
default:markdown

markdown returns the page as Markdown, or the AI's answer as text when you send a prompt. json returns structured data: on popular sites with no prompt, the page's data; elsewhere, data the AI extracts. Lowercase only.

Available options:
markdown,
json
scrapeMode
enum<string>
default:default

default opens every page in a real browser. fast fetches the page directly and opens a browser only when the page needs one, which is quicker for static pages. fast doesn't run actions or take screenshots. crawlerMode is accepted as an older name for this field.

Available options:
default,
fast
useProxy
boolean
default:false

Turns on Stealth Mode, which sends the request through residential proxies. Billed by bandwidth, not credits. Not needed when you set proxyCountry. false keeps the proxy off, even with a country. When Stealth Mode can't be used because your proxy bandwidth ran out, the request runs without it and warnings says so.

proxyCountry
string

The country to read pages from, as a two-letter code such as us, gb or de, or eu or asia. Setting a country turns Stealth Mode on (billed by bandwidth), and the response's warnings says so, unless useProxy is false: then the country isn't used, and warnings says that instead. global means any country.

Example:

"gb"

cookies
string

Cookies from your own signed-in browser session, to read pages that need an account. Either name=value; name2=value2 or rows pasted from your browser's DevTools. Never stored. See Logged-in pages.

screenshot
boolean
default:false

Takes a screenshot of the visible part of each page.

fullPageScreenshot
boolean
default:false

Takes a screenshot of the whole page, top to bottom. You don't need to set screenshot as well.

extractContentOnly
boolean
default:false

Removes headers, footers, menus and other page furniture from the Markdown, leaving the main content.

Response

The job was queued. Poll GET /scrape/{jobId} for the result.

status
enum<string>
Available options:
queued
jobId
string<uuid>

The job's ID. Use it with GET /scrape/{jobId}.

message
string

Where to poll for the result.

deduplicated
boolean

true when the same URLs were sent in the last 5 seconds and this is that job's ID.

schema_warnings
string[]

Only when your schema uses keywords that are ignored, such as anyOf. The job still runs.