Fetch and Read

Amnt

Paid

Live

Pricing

< $0.01

USDC

Call this tool

Turn any public web page into clean markdown an agent can actually read. Strips scripts, navigation, sidebars, cookie banners and comment threads, resolves every link to an absolute URL, and returns the title, site name and publication date alongside the text. Handles JSON and plain text too. Use it after a search tool: search gives you links, this reads them. HOW TO CALL IT: POST {"url": "https://example.com/article"}. Options: "maxChars" (default 20000) and "includeLinks" (default true). YOU GET BACK the page as clean markdown, plus its title, site name and publication date. WHAT IT STRIPS: scripts, styles, navigation, headers, footers, sidebars, cookie banners and comment threads — the boilerplate that makes raw HTML expensive to feed to a model. USE IT AFTER A SEARCH TOOL. Search gives you links; this reads them. It handles JSON and plain text as well as HTML, and follows redirects. LIMITS: public https pages only, 2 MB per page, 25 seconds. It runs no JavaScript, so a page that renders entirely client-side returns little. No login, no paywall bypass, nothing stored.

agent

amnt

fetch read

data

Parameter	Type	Description
url*	string	The public https page to read.
maxChars	integer	Cut the markdown at this length. The reply says whether it was cut.
includeLinks	boolean	Keep [text](url) links in the markdown. Turn off for prose only.

Parameter	Type	Description
title	string	The page title.
finalUrl	string	The URL actually read, after redirects.
markdown	string	The page's main content as markdown, with navigation, scripts, styling and boilerplate removed.
siteName	string	The publishing site's name, from og:site_name or the hostname.
truncated	boolean	True when the page was longer than maxChars and the markdown was cut.
wordCount	integer	Words in the returned markdown.
contentType	string	What the server said it served.
description	string	The page's own meta description, when it has one.
publishedAt	string	The publication date the page declares, ISO 8601, when it declares one.

Tool ID

583

Chain

Base

Last Checked

Created

Updated

Endpoint

https://www.amnt.io/api/agent/amnt/fetch-and-read

< $0.01 USDC

Parameter

Type

Description

url*

string

The public https page to read.

maxChars

integer

Cut the markdown at this length. The reply says whether it was cut.

includeLinks

boolean

Keep [text](url) links in the markdown. Turn off for prose only.

Parameter

Type

Description

title

string

The page title.

finalUrl

string

The URL actually read, after redirects.

markdown

string

The page's main content as markdown, with navigation, scripts, styling and boilerplate removed.

siteName

string

The publishing site's name, from og:site_name or the hostname.

truncated

boolean

True when the page was longer than maxChars and the markdown was cut.

wordCount

integer

Words in the returned markdown.

contentType

string

What the server said it served.

description

string

The page's own meta description, when it has one.

publishedAt

string

The publication date the page declares, ISO 8601, when it declares one.