Fetch

Get a web page as Markdown, HTML or plain text.

GET/api/v1/web/fetch

Alpha: this endpoint is new and its response fields may still change.

Costs 1 credit per request; failed requests are refunded. Authentication, the search_metadata envelope and error codes are the same as for every endpoint (see the API reference). Try it in the playground.

Parameters

Page

Capture a public page and return its content.

Parameter	Type	Description
`url`required	string	Public HTTP(S) URL to capture, 1-8,192 characters. Include http:// or https://; bare domains are rejected. The host must be public, on port 80 or 443, without embedded credentials.Up to 8,192 characters.
`remove_overlay`	boolean	Remove modal and cookie-banner overlays from the page.
`detach_invisibles`	boolean	Drop display:none elements from the page.
`referer`	string	HTTP(S) URL sent as the Referer.Up to 8,192 characters.
`assert_status_code`	integer	Expected origin HTTP status, 100-599. A different status is rejected; otherwise a rendered error page is a successful capture.Range: 100 to 599.
`robots_txt`	string	User agent, 1-256 characters, to check against the site's robots.txt; disallowed URLs are rejected.Up to 256 characters.
`respond_with`	string	Output format: content (default; main content as Markdown, or the whole page when extraction keeps too little), markdown (whole page as Markdown), html, text, frontmatter (content plus front matter), or a '+' combination such as markdown+frontmatter. For screenshots use /api/web/screenshot.Up to 64 characters.
`with_links_summary`	string	true, false or all: add a links list to the response.
`with_images_summary`	boolean	Add an images list to the response.
`retain_media`	"none" \| "text" \| "link" \| "image" \| "html"	How video, audio and embedded video appear: none, text, link (default), image or html.
`preset`	"reader" \| "index" \| "research" \| "agent" \| "spider"	Option preset. Fills only the options the request leaves unset.
`no_gfm`	string	true, false or table: disable GitHub-flavored Markdown, or only its tables.
`keep_img_data_url`	boolean	Keep inline data: images in Markdown.
`engine`	"auto" \| "browser" \| "curl"	auto (default; lightweight HTTP first, a browser when needed), browser, or curl (no JavaScript).
`token_budget`	integer	Integer 1-10,000,000. Reject the page when its token count exceeds the budget.Range: 1 to 10000000.
`markdown_chunking`	"true" \| "h1" \| "h2" \| "h3" \| "h4" \| "h5" \| "structured" \| "s1" \| "s2" \| "s3" \| "s4" \| "s5"	true, h1-h5, structured or s1-s5: split the Markdown and return chunks.
`max_tokens`	integer	Integer 500-10,000,000. Trim content to this many tokens.Range: 500 to 10000000.
`base`	"initial" \| "final"	Resolve relative links against the requested URL (initial, default) or the final URL after redirects (final).

Extraction

Parameter	Type	Description
`retain_images`	"none" \| "all" \| "alt"	Images to keep: all (default), alt (alt text only) or none.
`retain_links`	"none" \| "all" \| "text" \| "gpt-oss"	Links to keep: all (default), text (link text only), none, or gpt-oss (citation format with a links summary).
`target_selector`	string	CSS selector to extract, 1-2,048 characters; POST also accepts a list of 1-16 selectors. Implies waiting for the selector. Match-all selectors are rejected.Up to 2,048 characters.
`remove_selector`	string	CSS selector removed before extraction, 1-2,048 characters; POST also accepts a list of 1-16 selectors.Up to 2,048 characters.
`with_iframe`	string	true, false or quoted: include iframe content before extraction.
`with_shadow_dom`	boolean	Include open shadow DOM content before extraction.

Browser

Parameter	Type	Description
`wait_for_selector`	string	CSS selector to wait for after navigation, 1-2,048 characters. POST also accepts a list of 1-16 selectors.Up to 2,048 characters.
`user_agent`	string	User-Agent for the page request, 1-1,024 characters without control characters. Defaults to the browser's own.Up to 1,024 characters.
`page_timeout`	integer	Seconds from 1 through 180 for page loading and browser operations. Cannot extend the total request budget set by timeout.Range: 1 to 180.
`locale`	string	Browser locale as a language tag, 2-64 characters. Defaults to the browser's own.Up to 64 characters.
`respond_timing`	"html" \| "visible-content" \| "mutation-idle" \| "resource-idle" \| "media-idle" \| "network-idle"	Page readiness condition to reach before responding.

Response

A JSON object with search_metadata (request id, credits used and remaining) and the result groups the page returns. Groups and fields depend on the content of each page, so treat every field as optional.

Examples

cURL

curl "https://serpfast.com/api/v1/web/fetch?url=https%3A%2F%2Fexample.com&respond_with=markdown" \
  -H "Authorization: Bearer $SERPFAST_KEY"

Node.js 18+

const params = new URLSearchParams({
  "url": "https://example.com",
  "respond_with": "markdown"
});
const res = await fetch(`https://serpfast.com/api/v1/web/fetch?${params}`, {
  headers: { Authorization: `Bearer ${process.env.SERPFAST_KEY}` },
});
const data = await res.json();
console.log(data);

Python

import os, requests

params = {
  "url": "https://example.com",
  "respond_with": "markdown"
}
res = requests.get(
    "https://serpfast.com/api/v1/web/fetch",
    params=params,
    headers={"Authorization": f"Bearer {os.environ['SERPFAST_KEY']}"},
    timeout=120,
)
print(res.json())