Drogasil Brazil Pharmacy, Health & Beauty Scraper

Scrape the full product catalogue of drogasil.com.br, one of Brazil's largest pharmacy chains: medicines, vitamins and supplements, health devices, beauty, daily care, baby and pet.

Search by keyword, browse a whole catalogue section, or paste links. Every record comes back flat, with the barcode, the current and pre-discount price in BRL, the multi-buy ("leve mais por menos") offer, live stock, images, per-variant rows, and the full Brazilian pharmacy regulatory block.

---

What you get

Always (listing fields, no extra reads)

Field	Description
productId, sku	The store's own product code
name, brand	Product title and brand
url	Canonical product address (tracking parameters stripped)
image, images	Cover image and gallery
price, currency	Current price, always BRL
packSize	Pack size as the store states it (10 Comprimidos, 3g, 100ml)
prescriptionRequired	Whether a prescription is needed
isGenericMedicine	Generic (genérico) medicine flag
isKit	Bundled multi-product kit
requiresRefrigeration	Cold-chain product (termolábil)
soldBy	drogasil (the pharmacy's own stock) or marketplace (a partner seller)
categories, categoryPath	Category trail, as a list and as one string
productGroup	The store's own product grouping (MEDICAMENTO, OTC, PERFUMARIA, ...)
pharmaceuticalForm	Dosage form when the store publishes it on the card
drugClassificationCode	The store's own medicine classification code
badges	Promotional labels shown on the card
variantCount, variantAttributes	How many variants exist, and this one's option values
isSponsored	The row was a sponsored placement in a keyword search
searchMode, scrapedAt	Which route found the product, and when

With Fetch product details on (one extra read per product, billed once)

Field	Description
ean	Barcode / GTIN-13
originalPrice, discountPercent, onSale	Pre-discount price and the reduction
multiBuyUnitPrice, multiBuyMinQuantity	The "leve mais por menos" offer: buy N or more, pay this per unit
pixPrice, pixDiscountPercent	The instant-payment (Pix) price and its extra discount
installments	Instalment plan rows (count, amount)
stockQuantity, inStock	Live stock the store reports for the product
images	The full media gallery, in the store's own order
description, shortDescription, longDescription	Product copy, flattened to plain text
usageInstructions	The store's detailed product description line
activeIngredient	Active ingredient (princípio ativo)
dosage	Strength, e.g. 1G, 50MG
manufacturer	Manufacturer (fabricante) when published
healthMinistryRegistration	The Brazilian health-ministry / ANVISA registration number (registro MS)
prescriptionCategory, prescriptionStripeCode	The prescription stripe (tarja) classification and its code
prescriptionRequired	Derived from the published stripe code
controlledMedication	Controlled-substance flag
psychotropicListCode	Psychotropic control-list code
therapeuticClassCode, therapeuticClassName	Therapeutic class (ATC-style code plus its name)
packageInsertUrl, packageInsertPage	The package insert (bula): the PDF and the store's own page
weightKg	Shipping weight
sellerName, benefitProgramAvailable	Marketplace seller name; whether a medicine-benefit programme applies
breadcrumb	The product page's own category trail
variants	One row per variant: sku, name, attributes, price, stockQuantity, inStock

In incremental mode

changeType (NEW / UPDATED / REAPPEARED / UNCHANGED / EXPIRED), changedFields, firstSeenAt, lastSeenAt.

---

Two modes

Search mode

Give keywords, catalogue sections, or both. Each keyword and each section is collected separately, up to maxItems products and maxPages result pages each.

{
  "mode": "search",
  "queries": ["dipirona"],
  "categories": ["medicamentos"],
  "maxItems": 50,
  "fetchDetails": true
}

Keyword searches and section browsing read the store through different doors. Catalogue sections and pasted links are read from the store's pages; a keyword search is read from the store's own search interface (see Connection below for why). Both return the same product rows, filters and totals, and both are walked forward page by page in the same way.

URL mode

Paste any mix of product pages, catalogue section pages and search result pages. Each entry is recognised by the page's own type, so you never have to say which is which. A link that already names a page number (?p=3) is honoured, so you get results from page 3 onwards and not from page 1.

{
  "mode": "url",
  "urls": [
    "https://www.drogasil.com.br/dipirona-genericos-1g-com-10-comprimidos.html",
    "https://www.drogasil.com.br/beleza/maquiagem/batom.html?p=2"
  ],
  "fetchDetails": true
}

---

Filters and sorting

Everything in the filter section applies in both modes: it narrows a keyword search, a browsed section, and a section or search link you pasted.

Input	Applied by	Notes
brands	the store	Spell the brand exactly as the store's own brand filter does
minPrice / maxPrice	the store	In BRL
soldBy	the store	Pharmacy's own stock, or marketplace partners
excludeKits	the store	Drop bundled kits
sortBy	the store	Relevance, price low to high, price high to low
prescriptionFilter	this actor	The store publishes the prescription class but offers no filter for it
inStockOnly	this actor	Live stock is only published on the product page, so this needs fetchDetails

The two actor-applied filters are checked against the products that come back, so a run using them may return fewer than maxItems.

If you paste a link that already carries its own filters, they are kept and the inputs above are added on top.

---

Recurring monitoring

Two separate features, both in the Resume & recurring updates section:

• resumeFromRunId continues ONE interrupted run. Paste the previous run or dataset ID and the products it already collected are skipped, and not billed again.
• incrementalMode is for a schedule. The actor remembers the previous run of the same search itself. The first run returns everything as NEW; later runs return only NEW, UPDATED and REAPPEARED. emitUnchanged and emitExpired are off by default because they return, and bill, extra rows.

State is kept separately per search, filter set and detail setting, so two different searches never share a baseline. Set stateKey to name a campaign, or to deliberately share one.

scrapedAt is the only field excluded from change detection. Every price field, and live stock, are included: a restock, a sell-out or a new multi-buy offer is a real change and is reported as UPDATED with the field named in changedFields.

---

What you are charged for

• Product result for each product returned.
• Detail enrichment, charged once per product returned when fetchDetails is on. It is never charged for a product that incremental mode suppressed, and never for a product whose page could not be read.
• Actor start, the platform's standard start event.

Turn fetchDetails off for a listing-only run: it is one read per result page instead of one read per product, and the surcharge is not charged at all.

---

Send results into your apps (MCP connectors)

Optionally pipe results into the apps you already use, as well as into the dataset. Authorize a connector under Apify, Settings, API & Integrations, then select it in mcpConnectors. Notion, Linear, Airtable and Apify are offered in the picker, and any other Model Context Protocol server you have authorized can be added by its URL.

For Notion, also set notionParentPageUrl and each product becomes a page under it. The other connectors get a best-effort write, or a digest when the connector exposes no suitable write tool.

Each connector receives a condensed, human-readable summary per product: the product name plus the key fields flattened to plain text, with nested objects collapsed to their main value and long lists trimmed. The complete record always stays in the Apify dataset. maxNotifyListings caps how many products are written per connector, defaults to 50, and never affects the dataset. Leaving mcpConnectors empty skips this entirely, and a connector that fails never fails the run.

---

Notes on this store, measured live

Brazil only, and there is deliberately no country input. The store serves one market from one catalogue: every page declares lang="pt-BR", every price is in BRL, there are no hreflang alternates and no locale or country switcher anywhere in the site. A country input would be a knob that changes nothing, so there is none. The sibling site drogaraia.com.br is a separate storefront and is not in scope for this actor.

Reviews are not available. This is a real finding about the store, not an omission. The store renders no review data on any server-rendered page: there is no rating, no review count and no review body in the product payload, and no aggregateRating in its structured data. The product page has an "Avaliações" (reviews) anchor, but the section behind it is filled in by a third-party widget entirely in the browser, and that widget is switched off for the two product groups that matter most here, medicines and over-the-counter medicines. Re-checked on 2026-08-28: the widget's own endpoint for this store also answers every request with an empty body, so even a run that rendered the page in a browser would have nothing to read. There is therefore no fetchReviews input, because it could only ever return nothing.

Store pickup versus delivery is not published per product. The store decides fulfilment at checkout from your postcode, so no product page states "available for pickup". What the store does publish, and what this actor returns, is soldBy (its own stock, which its branches fulfil, versus a marketplace partner that ships the order itself), requiresRefrigeration for cold-chain items, and the controlled-medication flags, which are the constraints that actually change how an order can be fulfilled.

Prices are national. The store renders one national price per product on its public pages; branch-level pricing is only resolved after a postcode is entered in a session, so this actor reports the public price.

Connection. The store turns away connections it does not recognise, so the run reads it through a connection built for exactly that kind of store, set up for you out of the box: you do not need to configure anything in the Connection section, and leaving Apify Proxy off entirely is likely to get the run refused. Any pool you select there is still honoured: it is tried on its own first, and it stays available as a fallback for the whole run. If the store declines a request the actor retries a bounded number of times and then reports the failure, rather than returning an empty result as if the search had simply found nothing.

Keyword search goes through the store's own search interface. Since 2026-09-04 the store's search page itself refuses automated readers on every connection (while catalogue sections, product pages and pasted links keep being served). A keyword search is therefore read from the same JSON interface the store's own website posts to when a shopper types a keyword; it returns the same products, filters, totals and page counts as the page did. Nothing changes in the input or the output.

A run that reads nothing at all tells you which of the three things went wrong, because each one needs a different next step from you:

what the run reports	what happened	what to do
"Enable Apify Proxy in the Connection section"	no connection was set up, so nothing was ever sent	turn Apify Proxy on, or supply your own exits
"the connection selected for this run was refused before any request left it"	a connection was set up, but your account is not allowed to use the option that was picked, so the store never saw the request	pick another option in the Connection section, or one your plan includes
"Apify Proxy was already enabled for this run, so this is the store blocking this specific request"	the connection worked and the store itself turned the request away	retry later, or use a different route (browse by categories, or paste a link)

The second and third cases used to be reported as the first one, which sent people off to enable a connection they already had.

Result counts. The store's own declared page ceiling is not the item count divided by the page size: a 2,543 product section declares 42 pages at 48 rows a page, and 21 pages when the same section is sorted by price. Both of the store's own signals are honoured, whichever ends the results first, and an empty page ends them too. So maxPages is a ceiling, not a promise: a section can finish earlier because the store says it has.

---

Local development

python3.12 -m venv .venv
.venv/bin/pip install -r requirements.txt
.venv/bin/python -m pytest tests -q
apify run --purge

Verification

Re-verified against the live source on 2026-08-31 (build 1.1): the batom catalogue section reports 2,547 products server-side (2,543 at the previous audit, live churn); the Ruby Rose brand facet narrows it to 237 exactly as the site's own facet island declares, a 10 to 20 BRL price band narrows it to 245, the "sold by Drogasil" facet to 104, and dropping kits on the vitaminas section narrows 52,716 to 52,591. An unknown brand returns zero rows rather than silently ignoring the filter. The keyword route composes with the filters ("dipirona" reports 155 results; brand facets read off the search surface narrow it to 2 or 3 all-brand rows), and the price sorts reorder server-side (first row 6.22 BRL ascending vs 27.15 unsorted vs 4,182 descending). Pasted links resolve: a product page parses with its full regulatory block, and a section link carrying ?p=2 continues from page 2.