Hugging Face Spaces Scraper - AI App Data avatar

Hugging Face Spaces Scraper - AI App Data

Pricing

from $1.60 / 1,000 results

Go to Apify Store
Hugging Face Spaces Scraper - AI App Data

Hugging Face Spaces Scraper - AI App Data

Scrape Hugging Face Spaces pages and extract AI app names, URLs, creators, descriptions and machine-learning app metadata.

Pricing

from $1.60 / 1,000 results

Rating

0.0

(0)

Developer

Ben

Ben

Maintained by Community

Actor stats

0

Bookmarked

2

Total users

1

Monthly active users

15 days ago

Last modified

Share

Hugging Face Spaces Scraper

Collect individual application cards from the public Hugging Face Spaces directory. The Actor exports the Space title, canonical directory link, card description and owner account, while excluding navigation such as /spaces/launch.

Use the resulting list to discover demos, build a research shortlist or retain dated observations of public application descriptions. The Actor reads directory cards; it does not run the applications or validate their model outputs.

Quick start

Run the existing three-Space public example, or supply:

{
"searchUrls": ["https://huggingface.co/spaces"],
"maxResults": 3
}

The Actor uses HTTP to read the initial public response. A qualifying item has the URL shape /spaces/<owner>/<space>. The title comes from the card's heading; its status banner, likes, emoji badge and update time are not concatenated into the title.

Inputs

FieldMeaning
searchUrlsPublic Spaces directory pages to read. The directory homepage is the default.
maxResultsMaximum saved records per supplied page, from 1 to 1,000.

Only cards exposed in each supplied response are collected. Raising the cap does not automatically scroll the directory, fetch another page or enumerate every public Space. Query URLs are useful only to the extent that Hugging Face includes the requested cards in its HTTP response.

Duplicates on the same page are removed by Space URL. A Space encountered on two input pages may appear twice, with different input provenance. Preserve those observations or deduplicate across them according to your own workflow.

Observed output

The following public card was observed on September 11, 2026:

{
"source": "huggingface-spaces-scraper",
"kind": "AI apps",
"index": 1,
"title": "H3 Acceleration Arena",
"company": "multimodalart",
"url": "https://huggingface.co/spaces/multimodalart/h3-acceleration-arena",
"description": "Blind A/B ranking of MiniMax-H3 acceleration variants",
"input_url": "https://huggingface.co/spaces",
"source_url": "https://huggingface.co/spaces"
}

company retains the existing field name for compatibility. On an HTML card it is the owner account segment of the Space URL, which can represent an individual, a team or an organization. It is not a verified legal company name.

description contains only the paragraph belonging to that card, capped at 1,500 characters. An absent paragraph becomes null. input_url identifies your requested page; source_url identifies its final URL after redirects. index is the one-based export position for that page. Qualifying structured records may also retain the source object in raw.

What the records do not establish

The export does not include normalized likes, hardware, runtime status, commit history or update timestamps. It also does not download a repository, model weights, training data or generated content. Opening a Space separately may require accepting terms or providing credentials; this directory scraper does not supply that access.

A card's featured placement is an editorial or ranking observation. It does not prove API availability, quality, licensing suitability, active users or commercial adoption. Check the individual repository and model documentation before integrating an application into a product.

Using the export

Download JSON or CSV from the completed dataset. For an application inventory, keep the Space URL as the stable source key and add the date of your observation. Grouping by company gives an account-level view, provided you retain that field's account semantics.

A description change can be useful to review, but disappearance from a bounded directory sample is not a deletion signal. The featured selection can change, or your cap can stop before the Space appears. Keep earlier records until a direct observation establishes the new state.

For scheduled use, reuse a saved Task with the same inputs and inspect failure logs before replacing an existing inventory. This Actor does not create a schedule or send notifications by itself.

Pricing and limits

The Free-tier result price remains $2.00 per 1,000 saved Spaces, plus the existing Actor start event. Progressive plan discounts remain unchanged. Check the Pricing tab for your tier and the start charge at your selected memory.

maxResults applies per input page. Set a maximum total charge for a bounded run. If a page fails to load, valid rows saved from other pages remain available, and the run reports the failure. Start charges can apply even when no result is saved.

Parser changes and support

The September 11 correction rejects navigation URLs, reads the actual Space heading and prevents descriptions from absorbing neighboring cards. Structured data is read before script removal. Input fields, dataset view and prices remain compatible.

For missing or incorrect cards, open an issue with the public directory URL, expected Space URL and affected field. A shared run can help diagnose a current layout change. Do not post tokens or private application data in public issues.

For a separate tool-directory comparison, see AITopTools Scraper. Similar names across directories are discovery candidates, not verified identity matches.