Google Maps Scraper / GoogleMapsCollectorPython · Protobuf · aiohttp

Google Maps,
read at the
source.

I built a scraper that turns a Google Maps search into clean business records: name, address, phone, website, rating, hours and coordinates.

Google Maps has no public API for this business data, so mine reads the internal protobuf endpoint directly, over plain HTTP. No browser.

The 500 recorded places of the run below, over an illustrative search grid.
Built and maintained by me, Youssef Nagy, from Arish, Egypt. Open source on GitHub and PyPI · hosted on the Apify Store

How it works.

From a search phrase to streamed records
The pipeline A search phrase becomes a grid of cells over the area. Each cell is one request to the internal endpoint, answered in protobuf wire format. The answer is decoded into typed fields and de-duplicated by place_id and hex_id, then enriched with hours, phone, website and reviews, and streamed out as records. Checkpoints and adaptive rate limiting run alongside. query restaurants in Los Angeles, CA grid over the area zoom sets cell size internal endpoint · protobuf over HTTP 1 · len 2 · varint 3 · message 1 · f64 2 · f64 decode · dedup namestring ratingfloat latitude, longitudefloat place_id, hex_idkey seen enrich · stream record hours phone website reviews, optional out, async checkpoints · adaptive rate limit
  1. A query becomes a gridI resolve the place in the query to an area and cut that area into a grid of cells. The zoom level sets how fine the grid is, so a whole city is covered cell by cell instead of stopping at the first page of results.
  2. The internal endpoint, directlyEach cell is one request to the same internal endpoint the Maps web app calls. It answers in protobuf’s wire format, not JSON. I call it over HTTP with aiohttp: no browser, no API key.
  3. Decoded, then de-duplicatedI decode the nested messages into typed fields: name, address, category, rating, review count, coordinates and place_id. Neighbouring cells overlap, so every place is de-duplicated by place_id and hex_id before it goes further.
  4. Enriched with detailsWith details on, each new place gets a second request for its hours, phone, website, amenities, plus code and closed status. Reviews are a separate pass you switch on when you need them.
  5. Streamed, checkpointed, pacedRecords stream out asynchronously as they are decoded. Progress is checkpointed so a long run resumes instead of starting over, and the request rate adapts as it goes.

The recorded run.

Recorded, not live

Recorded real run · 7 October 2026 · query “restaurants in Los Angeles, CA” · 500 of 1,500 rows shown

Provenance note in the data file:
Real output of the published actor, recorded by its author. Business listings are public information from Google Maps. Not live; ratings and hours are as of the run date.

I ran Google Maps Scraper with place details on and reviews off. Photos are left out of this page: they are Google-hosted image URLs, and I don’t re-serve them. “Open now” uses the current time in Los Angeles, because the recorded hours are local to Los Angeles.

Loading the recorded places…

Loading 500 recorded places…

Choose a place

Pick a dot on the map or a row in the list to see the full record.

Up and down arrow keys move through the list.

    Inputs.

    The actor’s input fields, and what this run used

    These are the real input fields of Google Maps Scraper on the Apify Store. The run above set four of them; everything else stayed at its default.

    The run used only search and place details. The other fields turn the same records into leads. Email finding visits each business’s own website for addresses. Social finding pulls its Facebook, Instagram, LinkedIn, X, YouTube, TikTok and WhatsApp links. People finding lists the named staff a site publishes, with their own email, phone or LinkedIn. Ad intelligence looks the business up in the public Meta Ad Library and in Google’s Ads Transparency Center. All four are experimental add-ons, off by default.

    Set in this runDefault

    What to search for

    • queryWhat you would type in the Maps search box, such as “dentists in Miami”. The location is detected from the phrase.Set in this run: “restaurants in Los Angeles, CA”
    • searchQueriesSeveral searches in one run, merged and de-duplicated by place.empty
    • startUrlsGoogle Maps place or search links to start from instead of a phrase.empty
    • customGeolocationAn exact area: a point and radius, or a GeoJSON polygon.empty

    Results and detail

    • maxResultsUpper limit on businesses; 0 collects everything found in the area.Set in this run: 1500
    • languageTwo-letter language code for the result text.en
    • zoomGrid fineness from 1 to 21; higher means smaller cells.13
    • includeDetailsThe second request per place: hours, phone, website, amenities.Set in this run: on
    • includePhotosUp to 10 photo URLs per place.on
    • includePlaceExtrasAmenities, plus code and permanently-closed status.on
    • includePopularTimesThe weekly busy histogram. Experimental: Google rarely includes it in the HTTP response.off

    Reviews

    • includeReviewsEach place’s Google reviews: author, rating, date, text and owner reply.Set in this run: off
    • reviewsLimitReviews per business when reviews are on.5
    • minReviewRatingKeep reviews at or above this many stars; 0 keeps all.0
    • reviewKeywordKeep reviews whose text contains a phrase.empty
    • reviewsSortNewest first, or Google’s most relevant order.newest

    Lead enrichment, experimental

    • includeEmailsVisits each business website and its contact pages for email addresses.off
    • includeSocialsSocial profile links found on each business website.off
    • emailOnlyKeep only businesses with an email address.off
    • socialOnlyKeep only businesses with a social profile.off
    • onlyWithWebsiteKeep only businesses with a real website.off
    • onlyWithoutWebsiteKeep only businesses without one.off
    • maxPagesPerSitePages read on each website during enrichment.4

    People and contacts, experimental

    • includePersonnelThe people a business’s own site names: name, title, and their published email, phone or LinkedIn.off
    • maxContactsPerBusinessMost contacts kept per business, highest confidence first.10
    • personnelMinConfidenceDrop contacts below this confidence tier.low
    • onlyWithPersonnelKeep only businesses with a named person.off

    Ad intelligence, experimental

    • includeMetaAdsWhether the business runs Facebook or Instagram ads, from the public Meta Ad Library.off
    • includeGoogleAdsWhether it runs Google ads, from Google’s Ads Transparency Center.off
    • onlyRunningAdsKeep only businesses that are advertising now.off
    • adCountryCountry for the ad lookups.US

    Proxy

    • proxyConfigurationHow requests reach Google.not shown
    • proxyUrlAn optional custom proxy instead.not shown

    Run it yourself.

    Hosted, or on your own machine

    Honest limits