Skip to main content
GET
Fetch a web page

Authorizations

x-api-key
string
header
required

API key from the API Console.

Query Parameters

url
string<uri>
required

Full http(s) URL of the page to fetch.

format
enum<string>
default:json

Response format: json (structured, default) or context (LLM-ready markdown).

Available options:
json,
context
effort
enum<string>

How thoroughly to retrieve the page: low favors speed, max spends the most time extracting content — useful for pages that are slow to load or render content client-side. Omit for the server's adaptive default. Invalid values return a 400 listing the valid tiers. Fetch failures (422, 503) are never billed, regardless of effort.

Available options:
low,
medium,
high,
max
maxContentLength
integer
default:100000

Maximum content characters to return (default 100000, maximum 200000).

Required range: 1 <= x <= 200000
query
string

Optional query string to scope extraction to. When set, the response includes query_extracts, query_snippet, and query_hash fields scoped to this query. With your own maxContentLength, a page longer than it returns the sections that answer the query instead of its opening (as markdown in content, with no separate markdown field, and truncated: true).

preset
enum<string>

Named defaults: app (ordinary defaults; direct calls) or agent (used by the MCP server and CLI; fills format=context). Any parameter you send wins.

Available options:
app,
agent
maxAge
integer

Freshness ceiling in seconds. A stored copy older than this is read again, fresh, and charged for that read as usual. If a fresh read is not possible in time, the stored copy is returned with a warning giving its age.

Required range: x >= 0
ask
string

Up to 5 yes/no questions about the page (repeat the parameter or separate with |; up to 200 characters each). Answers come back in decisions as {question, p}. The answering call is part of the request's cost.

extract
string

Up to 10 field names (comma-separated or repeated). For each, fields gives the passage on the page that states it, or null. The answering call is part of the request's cost.

Response

Extracted page content

url
string
title
string
description
string
content
string

Clean extracted text content

markdown
string

Clean extracted content as markdown, when available.

extracts
string[]

Key passage extracts from the page, when available.

snippet
string

Short excerpt from the page.

author
string
date
string
site_name
string
image
string
lang
string
word_count
integer
truncated
boolean
query_extracts
string[]

Text extracts from the page most relevant to query. Only present when query is set.

query_snippet
string

Short excerpt from the page most relevant to query. Only present when query is set.

query_hash
string

Hash identifying the query used to scope this response. Only present when query is set.

decisions
object[]

Answers to ask questions: the probability of yes for each. Only present when ask is set and the answers arrived in time.

fields
object

For each extract field name, the passage on the page that states it, or null. Only present when extract is set and the answers arrived in time.

partial
boolean

Whether this is a partial response returned before the page finished retrieving. Only present when true.

retry_after_seconds
integer

Seconds to wait before retrying for the full content. Only present on partial responses.

metrics
object
body_source
enum<string>

Present only when the live page is gone or unreachable and this body is an archived copy. wayback = the Internet Archive's Wayback Machine.

Available options:
wayback
body_captured_at
string<date-time>

When the archived copy was saved. Present only with body_source.

body_snapshot_url
string

Link to the archived copy. Present only with body_source.