Mercado Livre / Mercado Libre Scraper

Scrape products from Mercado Livre (Brazil) and Mercado Libre (Argentina, Mexico, Chile, Colombia, Uruguay) — by search keyword, listing/category/seller URL, or direct product URL — and get back a rich, structured record for every product: price, original price, discount, installments, seller + reputation, rating, reviews, sold quantity, images, full specs and more.

No login. No cookies. No API key. Public data only — the client just runs it.

What you get (fields)

Every product row includes, when present on the page:

Field	Description
itemId	Mercado Livre id (e.g. MLB1046215784)
title	Product title
price / priceText / currency	Current price (number + raw + symbol)
originalPrice / discountPercent	List price before discount, and the % off
installments	Installment plan text (e.g. 10x R$335,48 sem juros)
pixInfo	Pix price/discount line (product pages)
seller / sellerReputation / officialStore	Store name, reputation line, official-store flag
rating / reviewsCount	Average star rating and number of reviews
soldQuantity	Units sold (e.g. 10 mil)
condition	New / used
availability	Stock / fulfillment status
freeShipping / fullShipping / shippingText	Shipping flags + raw shipping text
category	Breadcrumb category path (product pages)
pictures[]	Up to 25 full-size image URLs (product pages)
attributes[]	Full specification table {name, value} (product pages)
description	Full product description text (product pages)
tags[]	Promo labels (OFERTA DO DIA, MAIS VENDIDO, X% OFF, …)
url / permalink	Clean canonical URL + the raw link (ads resolved to a clean /p/ URL)
isAd	Whether the card was a sponsored result
position	Rank within the search results

Input

{
  "search": ["iphone 15"],
  "country": "BR",
  "sort": "price_asc",
  "condition": "new",
  "minPrice": 1000,
  "maxPrice": 5000,
  "freeShippingOnly": true,
  "maxItems": 100,
  "itemUrls": ["https://www.mercadolivre.com.br/produto/p/MLB123456"],
  "startUrls": ["https://www.mercadolivre.com.br/ofertas?category=MLB1051"]
}

• search — keywords/phrases, each opened as its own results page on the chosen country's storefront and paginated automatically up to maxItems.
• country (default BR) — BR, AR, MX, CL, CO, UY. Only affects search.
• sort — relevance (default), price_asc, price_desc. (search only)
• condition — all (default), new, used. (search only)
• minPrice / maxPrice — price-range filter in the country's currency. (search only)
• freeShippingOnly — restrict to free-shipping products. (search only)
• startUrls — arbitrary listing/category/seller/pre-filtered-search URLs, crawled and paged as result pages.
• itemUrls — full product permalinks, opened and scraped in full detail (always, regardless of maxItems).
• maxItems (default 50, max 2000) — cap on products from search + startUrls combined.
• proxyConfiguration (default: Apify RESIDENTIAL) — kept on RESIDENTIAL internally since that is what gets past the site's anti-bot.
• timeoutSecs (default 45) — per-page load timeout.
• capsolverApiKey (optional) — solve a sitekey captcha (Turnstile/hCaptcha/reCAPTCHA) via CapSolver if one is ever shown.

At least one of search, startUrls or itemUrls is required.

How it gets the data (reliably, and cheaply)

Mercado Livre's public REST API is now locked behind authentication (it returns 403 forbidden to anonymous callers), and the storefront's search (lista./listado.) and product (/p/…) pages answer a cold HTTP request — or a naive headless browser — with a "Segurança" / account- verification wall (their in-house suspicious-traffic device check). This actor gets through the way a real shopper does:

1. A real headful Chrome session (Playwright/Crawlee) on an Apify residential IP, with a consistent desktop fingerprint and human-like mouse/scroll activity.
2. It warms up on the homepage first so the site's security SDK establishes device trust, then reaches results by typing into the storefront's own search box (an in-site navigation) rather than a cold deep-link — which is what trips the wall.
3. If a verification interstitial still appears, it lets the automatic device check clear itself (and solves a press-and-hold challenge if one is shown) before retrying on a fresh residential session — a clean draw then passes straight through.
4. To keep cost down, images, media and fonts are blocked at the network layer, so each page only pulls the HTML it needs (far less residential-proxy bandwidth).

If a product genuinely cannot be loaded after all retries, the row carries an honest error field describing what was seen — never fabricated data. Error rows are not charged for.

Use cases

• Price & discount monitoring across BR/AR/MX/CL/CO/UY — no login or token.
• Competitor / seller / catalog research and BI pipelines.
• Deal and stock tracking.

⭐ Enjoying this Actor?

A quick rating/review helps others find it.