Developer and Reference Data API
Packages, repositories, models, papers, Q&A, encyclopedia articles and device specs from 42 developer and reference sites.
A developer and reference data API for the sites engineers and researchers read every day. Package registries: npm (details, versions, download counts, dependents and security advisories), PyPI (releases, dependencies and download stats) and Docker Hub (images and tags). Code and AI: GitLab projects, merge requests and pipelines, and Hugging Face models, datasets, Spaces, papers and the Open LLM Leaderboard. Q&A and communities: Stack Overflow, Hacker News, DEV Community and LeetCode. Reference and research: Wikipedia articles, page views and revision history, Fandom wikis, arXiv papers, the Wayback Machine, WordPress.org plugins and GSMArena phone specs. One key and one JSON shape instead of a separate client and rate limit per site.
- 42 platforms
- 385 endpoints
- from $1 / 1,000 calls
- 1,000 free credits / month
Developer & reference APIs on API.SH
Each platform has its own page with endpoints, example responses and prices. One API key works for all of them.
What teams build with developer & reference data
Open-source adoption tracking
Chart a package's npm or PyPI downloads, new versions and dependents over time, and compare it with alternatives before choosing a dependency or reporting on your own project's reach.
Developer community monitoring
Watch Stack Overflow questions for your product's tag, Hacker News stories and comments that mention it, and DEV articles, to answer users and spot recurring problems.
AI model and dataset research
Search Hugging Face models and datasets by task, read model cards, files and commit history, and track the Open LLM Leaderboard and Daily Papers.
Knowledge for RAG and agents
Give an AI agent Wikipedia articles with infoboxes and sections and arXiv papers as clean JSON, so answers can cite a source.
Research trend analysis
Combine Wikipedia page views, Hugging Face downloads and arXiv submissions in a category to measure attention to a topic over time.
Site history and changes
Find a page's Wayback Machine captures between two dates or the capture closest to a date, and read the old page to see what a site said before.
Popular developer & reference endpoints
| platform | endpoint | per 1,000 calls |
|---|---|---|
| Wikipedia | Wikipedia article (summary, infobox, sections, links) /wikipedia/ | $1 1 credit |
| Stack Overflow | Stack Overflow user profile /stackoverflow/ | $1 1 credit |
| Hugging Face | Hugging Face model search /huggingface/ | $1 1 credit |
| npm | npm package details /npm/ | $1 1 credit |
| PyPI | PyPI package details /pypi/ | $1 1 credit |
| GitLab | GitLab project /gitlab/ | $1 1 credit |
| Hacker News | Hacker News item with comments /hackernews/ | $1 1 credit |
| arXiv | One arXiv paper /arxiv/ | $1 1 credit |
| Docker Hub | Docker Hub repository /dockerhub/ | $1 1 credit |
| LeetCode | LeetCode user profile /leetcode/ | $1 1 credit |
| Wayback Machine | Wayback Machine capture closest to a date /wayback/ | $1 1 credit |
| Kaggle | Kaggle competition details /kaggle/ | $1 1 credit |
One request, clean JSON
Wikipedia article (summary, infobox, sections, links) from Wikipedia: GET https://api.api.sh/wikipedia/
curl "https://api.api.sh/wikipedia/v1/page?title=Albert+Einstein&lang=en" \
-H "x-api-key: $APISH_KEY" import os, requests
res = requests.get(
"https://api.api.sh/wikipedia/v1/page",
params={"title": "Albert Einstein", "lang": "en"},
headers={"x-api-key": os.environ["APISH_KEY"]},
timeout=120,
)
res.raise_for_status()
data = res.json() {"title": "Eiffel Tower","page_id": 9232,"lang": "en","url": "https://en.wikipedia.org/wiki/Eiffel_Tower","description": "Tower in Paris, France","summary": "The Eiffel Tower ( EYE-fəl; French: Tour Eiffel [tuʁ ɛfɛl] ) is a lattice tower on the Champ de Mars in Paris, France. It is named after th…","wikidata_id": "Q243","short_description": "Tower in Paris, France","disambiguation": false,"redirected_from": null,"thumbnail": "https://thumb.wikimedia.org/wikipedia/commons/thumb/8/85/Tour_Eiffel_Wikimedia_Commons_%28cropped%29.jpg/960px-Tour_Eiffel_Wikimedia_Commons…","image": "https://upload.wikimedia.org/wikipedia/commons/8/85/Tour_Eiffel_Wikimedia_Commons_%28cropped%29.jpg?utm_source=en.wikipedia.org&utm_campaign…","coordinates": {"lat": 48.85822222,"lon": 2.2945,"type": null,"name": null,"country": "FR","region": "75","dim": "1000","globe": "earth"},"last_modified": "2026-11-03T05:05:02Z","length": 97417,"last_revision_id": 1383208616,"last_edited": "2026-11-01T15:22:00Z","page_views": {"days": 59,"total": 254680,"daily_average": 4317,"daily": [{"date": "2026-09-05","views": 2925},{"date": "2026-09-06","views": 3290},"… 58 more"]},"infobox_type": "building","infobox": {"name": "Eiffel Tower","native_name": "Tour Eiffel","logo": "Eiffel Tower logo.svg","logo_size": "120","image": "Tour Eiffel Wikimedia Commons (cropped).jpg","caption": "Seen from the Champ de Mars, 2009","highest_start": "1889","highest_end": "1930","highest_prev": "Washington Monument","highest_next": "Chrysler Building","location": "7th arrondissement of Paris, France","coordinates": "48|51|29.6|N|2|17|40.2|E|region:FR-75","mapframe-zoom": "16","construction_start_date": "1887-01-28","completion_date": "1889-03-31","opening_date": "1889-05-15","building_type": "Observation tower; Broadcasting tower","architectural": "300 m","tip": "330 m","top_floor": "276 m","floor_count": "4 (3 accessible, 1 residual)","elevator_count": "8","architect": "Stephen Sauvestre","structural_engineer": "Maurice Koechlin; Émile Nouguier","main_contractor": "Compagnie des Etablissements Eiffel","owner": "City of Paris, France","operator": "Société d'Exploitation de la Tour Eiffel (SETE)","website": "https://www.toureiffel.paris/en","references": "I. talleststatus"},"sections": [{"number": "1","level": 2,"title": "History","anchor": "History","index": 1},{"number": "1.1","level": 3,"title": "Origin","anchor": "Origin","index": 2},"… 35 more"],"categories": ["1889 establishments in France","Architectural controversies","… 17 more"],"links": ["1900 Summer Olympics","1915 Çanakkale Bridge","… 498 more"],"links_truncated": true,"images": ["File:1000 Eiffel tower night illumination by Pierre Bideau in Paris Photo by Giles Laurent.jpg","File:2013 Fireworks on Eiffel Tower 11.jpg","… 51 more"],"languages": [{"lang": "af","title": "Eiffeltoring","url": "https://af.wikipedia.org/wiki/Eiffeltoring"},{"lang": "gsw","title": "Eiffelturm","url": "https://als.wikipedia.org/wiki/Eiffelturm"},"… 173 more"],"cached": false,"credits_used": 1} Why API.SH for developer & reference data
- Public data only: no logins, cookies or accounts on the source sites
- One key and one JSON shape across 42 developer & reference platforms and 918 in all
- Blocks, retries and parsing handled for you: no scraper to maintain
- Pay per call; failed calls cost nothing
- Also in n8n, Zapier, Make, Google Sheets and AI agents (MCP)
Compared with other data APIs
Other providers sell each platform as a separate actor, scraper or subscription. On API.SH every developer & reference platform is under one key and one balance, priced per call on the pricing page.
Developer & reference API questions
Why use API.SH instead of each site's own API?+
Many of these sites have good free APIs, each with its own client, key, rate limits and response format. API.SH puts them behind one key with consistent JSON and paging, and adds data some official APIs do not return, such as Stack Overflow pages or Hugging Face leaderboards.
Is the Wikipedia API free?+
Yes, Wikimedia's own APIs are free with a descriptive user agent and fair-use limits. API.SH's Wikipedia endpoints return articles with infobox and sections parsed, page views and revision history in the same format as everything else, using your 1,000 free monthly credits.
Do I need API keys for GitLab, npm or Hugging Face?+
No. API.SH reads public projects, packages, models and datasets without your tokens on those sites. Private repositories and gated models are not returned.
Can I get npm or PyPI download statistics?+
Yes. /npm/v1/downloads returns download counts for a package over a period, and the PyPI download statistics endpoint returns recent downloads, which you can store to chart adoption.
Is there a Stack Overflow API alternative?+
Stack Exchange's API is free with quotas per key. API.SH returns questions with answers and comments, search, tags and user profiles from Stack Overflow, plus related and linked questions, without managing a quota.
Is there a Google Scholar or research paper API here?+
This category has arXiv; Google Scholar is in the search category, and scholarly works search by DOI, ORCID and citations is in the public data category.
Can I use it from Python?+
Yes. Each endpoint is a plain GET request with your key in a header, so Python requests or httpx work directly; there is also a Python SDK and a JavaScript SDK.
Is there an MCP server for AI agents?+
Yes. The MCP server at api.sh/mcp exposes these endpoints as tools, so Claude Desktop, Cursor and other MCP clients can search Wikipedia, read a package or open a paper during a conversation.
How do I start?+
Get a free key, then call /wikipedia/v1/page with a title or /npm/v1/package with a package name. The docs at api.sh/docs show each endpoint's parameters.