Hugging Face API — Models, Datasets & Research Papers Now Available

Anysite now has 7 dedicated Hugging Face endpoints — covering models, datasets, and research papers — available via REST API, CLI, and MCP.

Hugging Face hosts over 900,000 models and 200,000 datasets. It is also the primary index for daily AI research: every significant arXiv paper is tracked on the hub with linked models, datasets, and Spaces that implement its results. Getting programmatic access to that catalog previously meant dealing with HF’s own authentication, rate limits, and deeply nested responses.

These endpoints remove that friction. Pass a repository id, a search query, or an arXiv paper id in a POST request. Get back structured JSON. No Hugging Face account required.

What is available

The 7 endpoints span three categories:

Datasets — look up a specific dataset by repository id or URL (/api/huggingface/datasets), or search the full catalog by keyword, author, and tag filters (/api/huggingface/datasets/search). Responses include download counts, trending scores, license, tags, card metadata, and the full file list.

Models — retrieve a model’s details by its repository id or URL (/api/huggingface/models), or search by keyword and filter on pipeline task, library (transformers, diffusers, etc.), language, or license (/api/huggingface/models/search). Responses include architecture config, safetensors info, and linked Spaces.

Papers — three endpoints: get a specific paper by arXiv id with linked models and datasets (/api/huggingface/papers), search papers by keyword (/api/huggingface/papers/search), and pull the day’s trending papers list (/api/huggingface/papers/trending). All paper responses include an AI-generated summary, author list, upvote and comment counts, and the arrays of linked artifacts.

Example request

Search for Llama-family text-generation models:

curl -X POST https://api.anysite.io/api/huggingface/models/search \
  -H "access-token: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "search": "llama",
    "filter": ["text-generation"],
    "count": 5
  }'

Pull today’s trending AI papers:

curl -X POST https://api.anysite.io/api/huggingface/papers/trending \
  -H "access-token: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"count": 10}'

Works inside Claude, Cursor, and ChatGPT

All Hugging Face endpoints are available on the MCP Unlimited plan ($30/mo). That means you can search models, look up datasets, and pull trending papers directly from your AI assistant — no REST client required. Ask Claude to find the top text-to-image models with more than 50,000 downloads, or to summarize this week’s most-upvoted research papers.

Getting started

Full endpoint reference, parameters, and response schemas are on the Hugging Face endpoint page. API documentation is at docs.anysite.io. Every Anysite plan includes access — start a free trial to try it out.