Hugging Face API: models, datasets, Spaces & papers

Why Social Fetch

Hugging Face data as JSON. You build the product.

Live Hugging Face data in one response shape, with platform changes handled for you.

Hugging Face, plus every other platform

One key and one response envelope for Hugging Face and 56 more platforms.

Pricing you can reason about

One request, one credit. Know how far your balance will go before you run your next batch.

Live data on every request

Fetch public data when you ask for it.

Nothing to maintain

We handle platform changes and breakages.

Built for production jobs

Run concurrent workloads with explicit statuses and request IDs.

Regular jobs. Better rates.

Running lookups every month? Get lower per-credit rates with a monthly plan. For occasional projects, choose pay as you go.

What Hugging Face data can you get?

42 endpoints, one API key. Every response comes back as live JSON in the same envelope.

Browse all 42 Hugging Face endpoints

Models, datasets & Spaces

Repository content

Accounts

Papers

Collections & feeds

Search & discovery

Dataset rows

Kernels

Use cases

What teams build with Hugging Face data

Monitoring, enrichment, research and content workflows on Hugging Face data, on the same API as every other platform.

What you can build

Track new models

Watch trending repositories, recent changes and daily papers.

Compare the catalogue

Rank models and datasets by downloads, likes, licence and hosted prices.

Browse all workflows →

How it works

Your first request in three steps.

Use TypeScript or your automation tool of choice.

Sources

+ 51 more platforms

One request

Social Fetch API

GET/v1/tiktok/profiles/charlidamelio

x-api-key: sk_live_••••••••

1 request · 1 credit

Same JSON shape

response.json200 OK

{
  "data": {
    "lookupStatus": "found",
    "profile": {
      "platform": "tiktok",
      "handle": "charlidamelio",
      "displayName": "charli d’amelio",
      "verified": true
    },
    "metrics": {
      "followers": 160263648
    }
  }
}

Step 1
Create your free account
Get an API key with 100 requests included.
Step 2
Choose an endpoint
Pass a username, URL, keyword, or post ID.
Step 3
Get clean JSON
Use the response immediately or send it into your pipeline.

Works with your stack

Code, automate, or connect your AI.

Explore integrations

TypeScript
n8n
Make
Apify
MCP
AI clients

Pricing

Social data, priced simply.

Start with 100 free credits. Monthly plans give you predictable volume and lower per-request rates.

Starter

$29.00

18,000credits

$1.61 per 1k · $0.00161 per request

Top up from $1.61 / 1k if you run out
Up to 25 monitors

Subscribe to Starter

Growth

$79.00

50,000credits

$1.58 per 1k · $0.00158 per request

Top up from $1.58 / 1k if you run out
5 team seats
Up to 100 monitors

Subscribe to Growth

Scale

$225.00

150,000credits

$1.50 per 1k · $0.00150 per request

Top up from $1.50 / 1k if you run out
25 team seats
Up to 250 monitors

Subscribe to Scale

Cancel anytime
30-day money-back guarantee
Price locked while you stay

Need 1M+ credits, invoicing, or volume rates? Talk to us

DevHunt

Product Hunt

Uneed

Customers

Teams building on Social Fetch.

“Our TikTok and Instagram scrapers broke constantly. Social Fetch took us off that rotation — we ship product again.”
Daniel Reeve · Founder, Social Crow

“Transcripts at volume from one endpoint. No broken scrapers for our small team to maintain.”
Marcus Lin · Founder, Voice Gecko
“Live profile data on every request. We never flag accounts from a stale cache.”
Hannah Webb · Founder, Scam Finder
“Heavy concurrent load, no rate limits to negotiate. The rare API I haven't wrapped in retries.”
Owen Pritchard · Founder, IP Flare

Hugging Face API: common questions

Is there a Hugging Face API for models, datasets and Spaces?

Yes. Social Fetch has routes for the model, dataset and Space catalogue: search with filters, one repository with its card metadata, the README, the file tree, commits, discussions, likers and accounts. Authenticate with your Social Fetch API key in the x-api-key header.

How do I search models by task or size?

Use GET /v1/hugging-face/models with pipelineTag set to a task such as text-generation, and minParameters or maxParameters set to values such as 500M or 7B. List the valid task ids with GET /v1/hugging-face/tags?group=pipeline_tag.

What does a model or dataset response include?

On found, a model carries the summary fields (id, url, author, likes, downloads, tags, licence, gated) plus card metadata, safetensors and GGUF facts, base models, evaluation results and hosted inference providers. A dataset carries its description, citation, languages, data sizes and card. include=config, spaces or security adds those model blocks.

How do I read a repository README or its files?

GET /v1/hugging-face/repos/{repoType}/{owner}/{name}/readme returns the full card as markdown, including its YAML header. The /files route returns the tree with sizes, git ids, large-file details and optional last commits. repoType is models, datasets or spaces.

Can I read dataset rows and statistics?

Yes. The /viewer route says whether rows can be shown, which subsets and splits exist and why rows are unavailable. The /rows route returns pages of typed values with page numbers, and the /statistics route returns per-column counts, ranges, histograms or label frequencies. Not every dataset has readable rows.

How deep can full-text search go?

GET /v1/hugging-face/search/readmes searches repository files, mostly READMEs, and returns a snippet of about 300 characters either side of the match. The deepest page starts after 500 skipped hits, because the platform slows down for each skipped hit and gives up after about 12 seconds.

How do I follow papers and collections?

GET /v1/hugging-face/papers/daily takes one of a day, an ISO week or a month; sending more than one is a 400. GET /papers searches by keyword, /papers/{arxivId} returns authors, links and citing repositories, and /papers/{arxivId}/content returns the full text. GET /collections searches curated lists and /collections/{owner}/{slug} returns every item in order.

Does HTTP 200 mean the repository or account exists?

No. HTTP 200 means the request finished. Check data.lookupStatus: not_found means no public repository, account or paper is visible under that name, and restricted means a repository exists but its content is gated. A private repository also returns not_found.

How much does the Hugging Face API cost?

Every route charges one credit per completed request, and each page of a list is billed separately. Check meta.creditsCharged on every response. New accounts start with 100 free credits.

Can I use Hugging Face routes from agents or MCP?

Yes. The routes are typed MCP tools and are reachable through Ask (POST /v1/ask) in plain English. Use the typed routes for production jobs.

The Hugging Face API in detailHow the Hugging Face routes work, what they cost, and how they differ from the official API.

Why teams use Social Fetch for Hugging Face

Social Fetch exposes 42 routes under /v1/hugging-face. The core routes cover the model, dataset and Space catalogue: GET /v1/hugging-face/models, /datasets and /spaces search with keyword, author, tag, licence and size filters, and the detail routes return one repository with its card metadata, likes, downloads and commit. Repository routes add the README as markdown, the file tree with optional last commits, the commit history, discussions and pull requests with their timelines, and the accounts that liked a repository.

Account routes cover a user or organization, the repositories it owns, followers and following, likes, recent activity, organization members, public storage buckets and profile details. Paper routes cover the daily list, search, one paper with its authors and citing repositories, the citing models, datasets and Spaces, and the full text as markdown. The rest of the catalogue adds collections, the social posts feed, the official and community blog, trending repositories, a site-wide search that groups results by kind, full-text search inside repository files, semantic Space search, tag values, hosted inference models with prices, recently changed pages, and dataset rows with their information and per-column statistics.

Every response uses the { data, meta } envelope with data.lookupStatus, so a missing or private repository, account or paper comes back as a typed not_found instead of an error. Lists page with an opaque cursor or a page number, depending on the route, and meta.creditsCharged is the billing source of truth.

How to call the Hugging Face API

1. Create an API key
Sign up at Social Fetch and copy a key (sfk_…). New accounts get 100 free credits, enough to try a search, a repository detail, a README and a paper lookup.
2. Pick the route
Use /v1/hugging-face/models, /datasets or /spaces to search, the matching /{owner}/{name} route for one repository, /repos/{repoType}/{owner}/{name}/readme for its card, and /papers/{arxivId} for a paper. Send x-api-key on every request.
3. Branch on the envelope, then bill
Check data.lookupStatus before storing a row (found, not_found or restricted), keep requesting with page.nextCursor while page.hasMore is true, and bill from meta.creditsCharged.

What each Hugging Face route returns

Search models, datasets and Spaces by keyword, author, task, library, language, licence, tags, gated access and parameter count.
Get one repository with its card metadata, safetensors and GGUF facts, base models, evaluation results and hosted providers.
Read a repository README as markdown, its file tree with optional last commits, its commit history and its discussion timelines.
Follow accounts: profile, owned repositories, followers and following, likes, recent activity and organization members.
Work with papers: the daily list, search, authors, citing repositories and the full text as markdown.
Read dataset rows and per-column statistics, or check up front whether a dataset's rows can be shown and why not.

How much does the Hugging Face API cost?

Every Hugging Face route charges one credit per completed request, shown on each endpoint page. Each page of a list is its own request. Credits never expire on pay-as-you-go packs, and meta.creditsCharged on every response is the billing source of truth.

Social Fetch vs the official Hugging Face API

Hugging Face's data has one object shape per kind, and many views need several calls to compose: a repository plus its README, a paper plus its citing repositories, a dataset plus its row facts and rows. Social Fetch gives you one consistent { data, meta } envelope across models, datasets, Spaces, papers, collections and accounts, typed found, not_found and restricted outcomes, cursor and page paging where each route needs it, and full-text search that returns a matched snippet rather than whole files.

Tools and guides

Pull your first Hugging Face data.

Run a live Hugging Face request in the playground. 100 free credits, no card, JSON back in seconds.

Get 100 free credits

No credit cardSign up in 30 seconds withGoogleGitHub