Hugging Face Scraper | Hub Models & Authors

Hugging Face scraper (also searched as HuggingFace Hub scraper) for the public Hugging Face Hub — search and export models, datasets, Spaces, papers, and collections without an API token. Rows include author names, profile socials, and the Hugging Face org press email when published. There are no phones and no street addresses on guest Hub pages. A practical Hugging Face Hub API / model-card extractor when you need guest-visible metadata in CSV, Excel, or JSON, not model weights.

This Actor is built for no-code operators first: pick a search type, enter a query (or a repoId), set maxResults, click Start, and download the Dataset. Developers can call the same Actor from the Apify API, Make, n8n, Zapier, LangChain, or a scheduled task.

Thin Apify Actor + managed worker. Guest Hub REST only. The worker owns residential egress. You do not send a Hugging Face token, cookies, or proxyUrl. This Actor does not download weights or launch Spaces.

---

What does the Hugging Face Scraper do?

The Hugging Face Scraper collects public Hugging Face Hub listings and detail metadata that a logged-out visitor can already see. It talks to Hub REST (/api/models, /api/datasets, /api/spaces, /api/papers, /api/collections) — no Hugging Face account and no hfToken.

Give it a query and a searchType, or a live org/name / paper id. The Actor:

1. Validates the input (query, searchType, repoId when required)
2. Calls a managed worker that fetches huggingface.co guest JSON
3. Parses opened surfaces into normalized Dataset rows (id, URL, downloads, likes, pipeline tag, license, tags, person, emails, …)
4. Dedupes and caps output with maxResults / maxPages. enrichDetails defaults on so author/org rows get profile leads.
5. Writes INPUT_ECHO, RUN_SUMMARY, and (on failure) ERROR_SUMMARY to the key-value store

Opened surfaces (what actually works)

Mode (searchType)	What you scrape	Required input
models	Model search / catalog	query optional; empty lists the default Hub feed
datasets	Dataset search / catalog	query optional; modality, language, filterTags
spaces	Space search / catalog (REST; HTML pager not used)	query optional; spaceFilter (e.g. gradio, docker)
papers	Papers catalog	Empty query lists the default feed (search= is ignored by Hub REST)
collections	Collections catalog	Empty query lists the default feed (search= is ignored by Hub REST)
model / dataset / space	Repo detail metadata	repoId (org/name)
paper	Paper detail	repoId (arxiv-style id, e.g. 2608.14978)
collection	Collection detail	repoId (slug org/collection-id)
files	Files tree metadata (path, type, size)	repoId + optional repoKind

Optional filters: pipelineTag (models), filterTags (e.g. license:mit, diffusers), author, sort (downloads, likes, createdAt, lastModified). Optional detailUrls for live huggingface.co model / dataset / space / tree / paper / collection URLs.

Honest limits (read this before you buy a competitor)

This Actor stays on guest Hub metadata. That is the point — and the constraint.

• No phones and no street / 门牌. Hugging Face profiles do not publish a mailing address or phone. phones[] is empty with leadAbsences.reason=本站无. Model ids that look like dates are not phone numbers.
• Email is rare. press@huggingface.co is the Hugging Face org press mailbox (plain mailto, already visible). User profiles usually have no mailto. git@hf.co is a git clone address, not an outreach email — it is never written to emails[]. This Actor does not stamp press@ onto every model row.
• No Hugging Face token / no login. Opened guest surfaces work without hfToken. Gated repos stay gated flags; we do not store cookies or mint sessions.
• No weight download. We do not fetch /resolve blobs, LFS, or safetensors files. searchType=files is paths and sizes only.
• No Space launch / no Inference click. Space rows are listing + detail metadata. /inference/models is not a repo id.
• Papers / collections keyword search is ignored by Hub REST (measured). Those modes list the default catalog + detail. Use models / datasets / spaces when you need search=.
• HTML sort=trending is default Hub order. API sort=trending returns 400 — omit sort or use downloads / likes.
• Not opened: Hugging Face Enterprise Hub, Community / discussions / commits, Home shelf, login, Inference Providers, hf-mirror.com as a separate catalog.

Other Store Actors may add author-website emails, eval-result enrichment, or monitor-only-new-releases. This one does not. Use it when you want tokenless Hub listings + papers/collections + file-tree metadata, exported as tables, at $3.00 / 1,000 results with developer-paid platform usage.

---

What Hugging Face data can I extract?

Typical default-dataset fields (exact keys vary by mode):

Field	Example / notes
id	org/name, paper id, or collection slug
url	https://huggingface.co/... link
kind / type	model, dataset, space, paper, collection
pipelineTag	Hub task (text-generation, fill-mask, text-to-image, …)
downloads / likes	Popularity when the listing exposes them
license	From license: tags or model card
tags	Hub tags (library, language, modality, …)
authorName	Owner / org login before enrich; display name after enrichDetails
person.name	Profile H1 / og:title (sample: Emils Petracenoks on thoughtcast)
emails	Usually empty. press@huggingface.co only on Hugging Face org rows
web.socials	Profile X / GitHub when they match the author login — not site footer links
phones / address.street / geo	Always empty; see leadAbsences (本站无)
leadSource / leadReveal / leadAbsences	Where the lead came from, and why phones/street/email are empty
gated	Whether the repo is gated
files / fileCount	Tree metadata only (searchType=files) — no oid, no blobs

You get structured rows you can join, filter, and schedule — not a dump of model binaries and not a clone of the repo.

You do not get: weight files, /resolve URLs, Space runtime logs, Inference Provider responses, or anything behind Hugging Face login / Enterprise Hub.

---

Why scrape Hugging Face?

Hugging Face is the default public index of open models, datasets, Spaces, and ML papers. Scraping the Hugging Face Hub is useful when you need a spreadsheet or pipeline of downloads, likes, licenses, and pipeline tags without paging the website or maintaining a Hub client.

Common jobs:

• Model discovery — find models by keyword or pipelineTag (text-generation, text-to-image, automatic-speech-recognition) and export downloads + URLs
• License / stack audits — filterTags such as license:mit or diffusers; dataset language=zh via Hub filter=language:zh
• Dataset & Space catalogs — list datasets by modality (audio, image, text) and Spaces by SDK (gradio, docker, streamlit)
• Papers & collections — pull the current Hub papers feed and curated collections (catalog + detail)
• File inventories — searchType=files for path/size metadata without downloading weights
• Automation — feed CRM, MLOps dashboards, competitive intel, or an AI agent via CSV / JSON / the Apify Dataset API

Why not just use huggingface_hub? The official library is the right choice when you already have a token, need to download weights, or must manage private repos you own. This Actor is a Hugging Face API alternative for guest metadata: no token to rotate, export-ready tables, and pay-per-result pricing.

Why not a lead-gen Hub scraper? Those tools are stronger when you want author websites and contact emails. This Actor is the opposite trade: public, tokenless, listing/detail metadata only — no website crawl, no email enrichment.

How this Hugging Face scraper compares

Store search for “Hugging Face scraper” also surfaces model-only Actors and lead-gen Hub crawls. This one is the tokenless Hub catalog + papers/collections option.

	This Actor	Typical Store peers
Models, datasets, Spaces	Yes (one searchType per run)	Often models-only, or all three in one run
Papers & collections	Yes (catalog + detail)	Usually not
Files tree (path/size)	Yes — no weight download	Rare
Author emails / website crawl	No	Some lead-gen Actors
Eval results / GGUF / param counts	No	Some enrichment Actors
Hugging Face token	Not required	Usually not required
Price (FREE / BRONZE)	$3.00 / 1,000 + $0.005 start	About $2.20–$5.00 / 1,000

Use a peer when you need emails, eval tables, or GGUF flags. Use this Actor when you want Hugging Face Hub listings you can export today, including papers and collections.

---

Pricing

Pay per event. You pay for Actor start + each result written to the default dataset. Platform compute / memory for this Actor is paid by the developer (User pays platform usage costs = No).

Event	Price
Actor Start (apify-actor-start)	$0.005 per start (per GB of memory, minimum one event)
result (apify-default-dataset-item)	$3.00 / 1,000 on FREE & BRONZE ($0.003 each)

Volume examples (FREE / BRONZE, one start at 1024 MB)

Results	Event charges (approx.)
100	$0.005 + $0.30 = $0.305
1,000	$0.005 + $3.00 = $3.005
10,000	$0.005 + $30.00 = $30.005

Store volume tiers (per result)

Tier	$/1,000	per result
FREE	$3.00	$0.003
BRONZE	$3.00	$0.003
SILVER	$2.60	$0.0026
GOLD	$2.20	$0.0022
PLATINUM	$2.00	$0.002
DIAMOND	$1.60	$0.0016

Default memory is 1024 MB. Actor Start is charged per GB (minimum one event). Keep maxResults at 2–15 while you test a query.

Empty / failed runs that write zero default-dataset rows are not billed as results (you still pay Actor Start).

Free Apify-plan users: this Actor caps 10 runs and 200 results per run (developer limit, not Apify).

---

More Actors like this

Use a specialized Actor when one exists for your site — structured fields, better coverage, usually lower cost. Use Hugging Face Scraper for public Hub models, datasets, Spaces, papers, and collections. Use the general tools below when you need email/phone checks or a generic protected-page scrape.

General tools

• Email Risk Validator
• Phone Number Intelligence
• Stealth Web Scraper

Developer / Open source / AI

• Hugging Face Scraper
• GitHub Scraper
• Apple App Store Scraper
• Hacker News Scraper

Prefer a dedicated site scraper?

If your target has a specialized Actor below, use that first — usually cheaper and more structured than a generic URL scrape.

Jobs & Freelance

• LinkedIn Jobs Scraper
• Indeed Jobs Scraper
• Upwork Jobs Scraper
• Glassdoor Scraper
• Fiverr Gigs Scraper
• Bayt Jobs Scraper

E-commerce

• Walmart Product Scraper
• Amazon Search Scraper
• Shopee Search Scraper
• Etsy Scraper
• SHEIN Product Scraper
• Temu Product Scraper
• Target Product Scraper
• Allegro Scraper

Real Estate

• Zillow & Zumper Scraper
• Realtor.com Scraper
• Apartments.com Rental Scraper
• Rightmove Scraper
• Idealista Scraper
• realestate.com.au Scraper

Travel & Stays

• Booking.com Hotels Scraper
• Airbnb Listings Scraper
• Expedia Scraper
• TripAdvisor Scraper

Social & Content

• YouTube Research Scraper
• TikTok Scraper
• Reddit Scraper
• YouTube Shorts Scraper

Ads Intelligence

• Facebook Ad Library Scraper
• TikTok Ads Scraper

Local & Maps

• Google Maps Scraper

→ See the full catalog in Related Actors below, or browse apify.com/lentic_clockss.

---

How to scrape Hugging Face models

No code required.

1. Open Hugging Face Scraper in Apify Console (Actor lentic_clockss/huggingface-scraper).
2. Set Search type — start with models (or datasets / spaces / papers / collections).
3. Enter a Search query (for example whisper, qwen, llama) or leave it empty to list the default Hub catalog.
4. Optional: set pipelineTag (text-generation), filterTags (license:mit), dataset language (zh), or Space spaceFilter (docker).
5. Set maxResults to 5–15 for a first run and maxPages to 1.
6. Click Start. When the run finishes, open the Dataset tab.
7. Download CSV, Excel, or JSON.

That is enough for most spreadsheet and no-code workflows. Developers can skip the UI and POST the same JSON to the Apify Runs API.

Example: scrape Hugging Face models

{
  "query": "whisper",
  "searchType": "models",
  "maxResults": 15,
  "maxPages": 1
}

Example: filter models by pipeline tag

{
  "query": "",
  "searchType": "models",
  "pipelineTag": "text-to-image",
  "maxResults": 15,
  "maxPages": 1
}

Example: scrape Hugging Face datasets by language

{
  "query": "",
  "searchType": "datasets",
  "language": "zh",
  "maxResults": 10,
  "maxPages": 1
}

Example: Hugging Face Spaces by SDK

{
  "query": "",
  "searchType": "spaces",
  "spaceFilter": "gradio",
  "maxResults": 10
}

Example: model detail + files tree (metadata only)

{
  "searchType": "files",
  "repoId": "openai/whisper-tiny",
  "repoKind": "models",
  "maxResults": 1
}

---

Connect to your tools

Use this Actor from your existing stack:

Platform	How to connect
Make.com	Search "Apify" → "Run Actor" → Actor ID: lentic_clockss/huggingface-scraper
n8n	Add Apify node → "Run Actor" → same Actor ID
Zapier	Apify integration → "Run Actor"
LangChain	ApifyActorsTool("lentic_clockss/huggingface-scraper")
Python / Node.js	Apify SDK or direct HTTPS call

API call example

curl "https://api.apify.com/v2/acts/lentic_clockss~huggingface-scraper/runs" \
  -X POST \
  -H "Authorization: Bearer YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"query":"qwen","searchType":"models","maxResults":10,"maxPages":1}'

Read results from the Dataset API:

GET https://api.apify.com/v2/datasets/{datasetId}/items?format=json

You can also schedule the Actor (daily model watch, weekly dataset inventory) from Console.

---

FAQ

Phone and street/门牌: no — Hugging Face does not publish them on guest pages (leadAbsences.reason=本站无). Email: only the Hugging Face org press mailbox press@huggingface.co is a reliable public mailto; most authors have none. Leave enrichDetails on (default) to fill person.name and matching socials from /{author}.

Do I need a Hugging Face API token? No. Opened guest surfaces run without hfToken. Do not paste a token into input.

Is this a Hugging Face API alternative? For public Hub listing and detail metadata, yes: you export CSV / Excel / JSON without managing Hub clients or tokens. It is not a replacement when you need weight downloads, private repos, or Inference Providers.

Can I scrape models, datasets and Spaces together? Each run uses one searchType. Run three times (or three tasks) and merge Datasets if you need all three catalogs in one sheet.

Can I download model weights? No. This Actor does not clone repos or fetch /resolve blobs. Use huggingface-cli download or the official Hub library for source.

Why is papers search ignoring my query? Hub REST GET /api/papers?search= does not change the first results (measured). searchType=papers lists the default papers feed. Use paper detail with repoId for a known arxiv id.

Does sort=trending work? Not on the Hub REST list. Omit sort (default Hub order, closest to HTML trending) or use downloads / likes / createdAt / lastModified.

How is this different from other Hugging Face scrapers on Apify Store? ParseForge’s model scraper is models-only (~$5 / 1k). Scrape Sage adds eval results and monitor-new-releases. This Actor covers models + datasets + Spaces + papers + collections + files-tree metadata, emits author names / socials (phones and street are 本站无; press@huggingface.co only on the Hugging Face org), stays tokenless, does not download weights, and prices PPE at $3.00 / 1,000 results with developer-paid platform usage.

Do I need to bring proxies? No. Do not send proxyUrl. The worker provides egress (WORKER_PROVIDES_PROXY=1).

What if a run returns 0 rows? The query may be too tight, the filter tag may have no matches (filter=language:zh on models is not applied — language is datasets-only), or the mode may be out of scope (Enterprise / login). Retry with an opened searchType and a broader query.

Is it legal to scrape Hugging Face? You are responsible for how you use the data. This Actor only collects public guest-visible Hub metadata, does not log in, and does not download weights. Follow Hugging Face Terms and your own compliance review. Unofficial — not affiliated with Hugging Face, Inc.

Can I use it from Make, n8n, Zapier, or LangChain? Yes. Actor ID lentic_clockss/huggingface-scraper. See Connect to your tools.

---

Input reference

Field	Type	Description
query	string	Hub search= for models/datasets/spaces. Empty lists the default catalog. Ignored for papers/collections
searchType	string	models, datasets, spaces, papers, collections, model, dataset, space, paper, collection, files (opened). Others in the schema are not supported (enterprise, community, home, login, inference)
repoId	string	org/name for detail/files, paper id, or collection slug
owner / name	string	Combined when repoId is empty
repoKind	string	models, datasets, or spaces for files/detail when the URL is ambiguous
pipelineTag	string	Models pipeline_tag= (text-generation, fill-mask, text-to-image, …)
filterTags	array	Extra Hub filter= tags (license:mit, diffusers, safetensors). Do not pass text-classification here (use pipelineTag)
modality	string	Datasets filter=modality:{value} (text, audio, image)
language	string	Datasets only: filter=language:{code} (zh, en, ja). Not a country market
spaceFilter	string	Spaces filter= (gradio, docker, streamlit, static)
author	string	Optional Hub author=
sort	string	downloads, likes, createdAt, lastModified. HTML trending = omit sort
maxResults	integer	Cap emitted rows (schema 1–200)
maxPages	integer	Hub REST pages via Link cursor, hard cap 20
detailUrls	array	Optional huggingface.co model / dataset / space / tree / paper / collection URLs. /resolve and /inference/models are rejected
enrichDetails	boolean	Default true. Fetch public /{author} profile HTML (name, socials, mailto). Does not invent phones or street addresses
workerBaseUrl	string	Optional HTTPS override. Production source of truth is Actor env

Do not send proxyUrl, cookies, or a Hugging Face token.

Full output schema is on the Dataset tab. Key-value store: INPUT_ECHO, RUN_SUMMARY, ERROR_SUMMARY.

---

Related Actors

All 81 Actors from [lentic_clockss]. Click a name to open the Store detail page.

Developer / Open source / AI

• Hugging Face Scraper
• GitHub Scraper
• Apple App Store Scraper
• Hacker News Scraper
• Stealth Web Scraper
• Email Risk Validator
• Phone Number Intelligence

Jobs & Freelance

• LinkedIn Jobs Scraper
• Bayt Jobs Scraper
• Fiverr Gigs Scraper
• Freelancer.com Scraper
• Glassdoor Scraper
• Himalayas Jobs Scraper
• Indeed Jobs Scraper
• Jobicy Remote Jobs Scraper
• RemoteOK Jobs Scraper
• SEEK Jobs Scraper
• Upwork Jobs Scraper

Real Estate

• Zillow & Zumper Scraper
• Realtor.com Scraper
• 99.co Scraper
• Realtor.com Agents Scraper
• Apartments.com Rental Scraper
• Bayut Scraper
• Craigslist Housing Scraper
• Daft.ie Scraper
• Dot Property Scraper
• FINN.no Scraper
• Funda Scraper
• Hepsiemlak Scraper
• Idealista Scraper
• Immobiliare.it Scraper
• ImmoScout24 Scraper
• Naver Land Scraper
• OpenSooq Scraper
• Otodom Scraper
• Property Finder Scraper
• PropertyGuru Scraper
• realestate.com.au Scraper
• Realtor.ca Scraper
• Rightmove Scraper
• SeLoger Scraper
• SUUMO Scraper
• Zillow Group Scraper

E-commerce

• Shopee Search Scraper
• E-commerce Scraper
• 1688 Global Product Search Scraper
• Allegro Scraper
• Amazon Search Scraper
• ASOS Product Scraper
• Cdiscount Product Scraper
• Costco Product Scraper
• Coupang Product Scraper
• Etsy Scraper
• Lazada Scraper
• MercadoLibre Scraper
• Mercari Japan Scraper
• Rakuten Japan Scraper
• SHEIN Product Scraper
• Target Product Scraper
• Temu Product Scraper
• Walmart Product Scraper
• Watsons Product Scraper

Travel & Stays

• Booking.com & Airbnb Scraper
• Agoda Scraper
• Airbnb Listings Scraper
• Booking.com Hotels Scraper
• Despegar Scraper
• Expedia Scraper
• Traveloka Scraper
• Travelstart Flights Scraper
• Trip.com Scraper
• TripAdvisor Scraper

Social & Content

• TikTok Scraper
• Reddit Scraper
• YouTube Shorts Scraper
• YouTube Research Scraper

Ads Intelligence

• Facebook Ad Library Scraper
• Google Ads Transparency VN
• TikTok Ads Scraper

Local & Maps

• Google Maps Scraper

→ Browse the full profile: apify.com/lentic_clockss

---

Changelog

• 0.1.4 — README SEO: Hub API / library keywords, competitor comparison, Related Actors count.
• 0.1.3 — Store README (SEO, How to use, pricing, More Actors, Related Actors). PPE: Actor Start $0.005 + result $3.00/1k with volume tiers. Platform usage paid by developer.
• 0.1.2 — Hub expansion: filterTags, dataset filter=modality: / filter=language:, papers/collections catalog + detail, extra pipeline-tag matrix cells.
• 0.1.1 — First Cloud Run worker pairing; models/datasets/spaces search + detail + files-tree metadata.