GitHub Marketplace Scraper

Collect names, item URLs and card summaries from public GitHub Marketplace result pages. Use a dated export to compare listed tools or keep a review queue with a direct link back to each Marketplace item.

Start with the saved example

Open the public example to inspect its input and run results. Its input requests three records from one page:

{
  "searchUrls": [
    "https://github.com/marketplace?type=apps"
  ],
  "maxResults": 3
}

Run the example, open the dataset and inspect the returned URLs and text before increasing the limit. A small result cap keeps the first review easy to check. A successful run means the fetch and extraction completed; it does not certify any listed product or establish how many people use it.

What the Actor reads

Marketplace pages also contain category menus and a Create a new extension link. Those controls are not products. This Actor reads individual Marketplace item cards, takes the heading as the title and keeps the card summary separate from the App label. The supplied page controls which results are visible; the Actor does not follow pagination automatically.

The September 11 app-page sample starts with Render, LovableBot and GitGuardian. Their summaries are Marketplace wording, not an independent functionality or security assessment. Individual app and Action item paths can be recognized on supplied result pages. The saved example uses type=apps; it does not establish an exhaustive inventory of Actions or every Marketplace category.

The Actor requests each supplied page over HTTP and processes the response. It does not execute page JavaScript, scroll to load more results, sign in or follow each item for additional details. A browser may display more records after interactive loading. Raising the limit cannot expose records that are absent from the fetched response.

Input and limits

Field	Meaning
searchUrls	Public directory, search or category pages to request. The saved example above is the verified starting point.
maxResults	Maximum exported records per supplied URL, from 1 to 1,000. The schema default is 25.

The cap applies separately to each URL, not to the whole run. Two pages with a limit of three can produce up to six rows. Duplicate item links within one page are removed; a product appearing on two different input pages can appear twice, with distinct provenance. The Actor keeps the first accepted occurrence within a page. Ordering follows the response and is not a popularity score.

This is a page extractor, not a complete directory crawler. Supply additional supported pages explicitly when needed. Validate a new category or search URL with a small request because directories may redirect, localize content or change their layout. Source access and the available fields can change independently of the Actor.

Output example

This record comes from the September 11, 2026 source-page snapshot. Names and wording may change on later runs.

{
  "source": "github-marketplace-scraper",
  "kind": "developer apps",
  "index": 1,
  "title": "Render",
  "url": "https://github.com/marketplace/render",
  "description": "Continuous integration and deploys with Render, the modern cloud for ambitious developers",
  "input_url": "https://github.com/marketplace?type=apps",
  "source_url": "https://github.com/marketplace?type=apps"
}

Field	Meaning
title	Name on the individual item card.
url	Absolute item URL on the source site.
description	This card's summary, or null when it has none.
source	Actor source identifier for grouping exports.
kind	Existing source-type label, not a quality classification.
index	One-based position among the accepted rows for this input page.
input_url	The original requested URL.
source_url	The final fetched URL after redirects.
company	Optional source creator/publisher byline. It is not verified company ownership.
raw	Optional original record if an input response supplied structured JSON.

Optional fields may be absent. Keep null values when exporting JSON; converting them to empty strings can hide the difference between missing information and an actual source value. The default dataset and its typed output are available through the Output tab. The existing Overview table keeps the main item and provenance columns together.

Keep a useful review history

Download JSON when you need to preserve nulls and optional fields. CSV or Excel is convenient for a manual review queue. Store the dataset or run ID with each export, because the input and source URL alone do not identify its collection time. Keep an unchanged copy of each capture before adding internal notes or scores.

For repeat collection, save the input as an Apify Task and schedule it only after checking the output. Compare exact item URLs across dated exports. A missing item in a capped response is not evidence that the product was removed: ordering, source changes and a different result limit can all change the sample. Treat a disappearance as a review question until the original item page confirms it.

The Actor does not keep monitoring state, send notifications or merge old datasets. Your workflow controls retention, comparisons and delivery. Apify webhooks or integrations can hand a finished dataset to an existing workflow; verify its success status before treating the export as a complete capture. Do not overwrite a previous valid snapshot with an incomplete run.

Errors and partial output

A failed page fetch marks the run as failed. Rows collected successfully from earlier input URLs remain in the dataset. Check both the final run status and the server log when processing several pages. A nonempty dataset on a failed run is partial evidence, not confirmation that every requested page worked.

A page without accepted item cards can return no rows. Inspect the actual source page and its final URL before deciding whether that reflects an empty result or a changed layout. Navigation links are excluded even when this leaves the dataset empty. Report a reproducible layout problem through the Actor's Issues tab with the public input URL and run ID; exclude credentials and private account data.

Pricing

Charges apply to exported dataset rows plus the existing memory-based Actor start event. Current result rates per 1,000 rows are:

Apify plan	Result charge per 1,000
Free	$2.5
Bronze	$2.25
Silver	$2.125
Gold	$2
Platinum	$1.875
Diamond	$1.75

The Free-tier start event is $0.00005 per memory unit, with one unit per GB and a minimum of one. Paid tiers retain their existing discounts. A three-row run at 512 MB has one start event; the Actor's default memory remains 4,096 MB, which uses four start units. See the Pricing tab for current charges and platform billing details. Set a total charge cap when running through the API or Console.

Changing input URLs or asking for more results can increase the total number of exported rows. A low result cap does not mean the server only downloads that many source records. Owner verification runs and resource-usage receipts are not customer revenue or a guarantee of future operating cost.

Questions before using the data

Does the export include installation or security evidence? No. The Actor does not install apps or Actions, access a repository, inspect code or run a security assessment. Marketplace descriptions and item names are discovery information. Review the linked item, permissions, publisher and repository requirements separately before using a tool.

Can it read private workspaces or accounts? No. It reads public pages without an account session. Do not supply credentials or private workspace URLs.

Does it guarantee a complete category inventory? No. It reads the requested response pages and applies a per-page limit. There is no automatic pagination or completeness claim.

Can I use a schedule or API? Yes. Apify supports saved Tasks, schedules, webhooks and dataset APIs. Configure these in your account and check one bounded run before enabling recurring work.

How should I use source claims? Keep the original URL and collection time beside them. Respect the source site's terms and applicable rights when reusing descriptions. An extracted claim remains source wording, even when it appears in a structured dataset.

Keywords: GitHub Marketplace, developer apps, Marketplace item cards, app summaries, JSON export, CSV export.