AI Agents Accessible APIsKnowledge and research data

Knowledge and research data APIs and MCP servers for AI agents

Encyclopedias, academic papers, news, computational knowledge and up-to-date library documentation.

16 knowledge and research data APIs an AI agent can call: auth method and header, official MCP server (3), free tier (16), docs links. Machine-readable JSON included. Agents: fetch catalog/knowledge-research.json instead of parsing this page. Last checked 2026-09-24.

Ratings and usage (★) fetched: .

ServiceWhat an agent can doAuthMCP serverFree tierDetails (with sources)NotesProof (ratings, usage)
arXiv APISearch arXiv preprints by keyword, author, category or ID and retrieve metadata, abstracts and PDF links.none
no auth		communityfree, no keyBase URL: http://export.arxiv.org/api/query
Free plan: no_key, no card ↗
Rate limits: No more than 1 request every 3 seconds, single connection at a time, across all your machines; max_results slices of 2000, 30000 total ↗
Example call: search preprints by keyword: GET http://export.arxiv.org/api/query?search_query={query}&start=0&max_results={n} ↗
MCP tools: 12: search_papers, get_abstract, download_paper, list_papers, read_paper, get_paper_outline … ↗		Endpoint: http://export.arxiv.org/api/query, returns Atom XML. Terms of Use allow no more than one request every three seconds over a single connection. For bulk harvesting use OAI-PMH or bulk data instead.GitHub 3,172★ blazickjp/arxiv-mcp-server
Smithery 5,142 uses
Congress.gov APILets an agent look up U.S. bills, amendments, votes, members, committees, nominations and Congressional Record data from the Library of Congress.API key
api.data.gov key: query parameter ?api_key=<API_KEY> or header X-Api-Key: <API_KEY>		communityFree: government API, key issued instantly by email; 5,000 requests/hour.Base URL: https://api.congress.gov/v3
Free plan: 5,000 requests per hour, no card ↗
Rate limits: 5,000 requests per hour ↗
Example call: list bills of a Congress by type: GET https://api.congress.gov/v3/bill/{congress}/{billType}?format=json&api_key={API_KEY} ↗
MCP tools: 11: congressgov_bill_lookup, congressgov_enacted_laws, congressgov_member_lookup, congressgov_committee_lookup, congressgov_roll_votes, congressgov_senate_nominations … ↗		Official Library of Congress API (v3); usage guide and endpoint docs in github.com/LibraryOfCongress/api.congress.gov. Paginated JSON/XML, max 250 results per page.—
Context7 (library docs for agents)Fetch up-to-date, version-specific library documentation and code examples into a coding agent's context via MCP or REST.API key
Authorization: Bearer <ctx7sk_... key> (optional; works at low limits without key)		officialFree: 1,000 API calls per month with a free API key (low limits without key); Pro $10/seat/monthBase URL: https://context7.com/api/
Free plan: 1,000 API calls, per month ↗
Rate limits: Low limits without key; plan-based limits with key; 429 with Retry-After and RateLimit-* headers when exceeded ↗
Example call: fetch up-to-date documentation snippets for a library: GET https://context7.com/api/v2/context?libraryId={libraryId}&query={query} ↗
Unit price: Per API call: Free 1,000 calls/month; Pro $10/seat/month includes 5,000 calls per seat, then $10 per 1,000 calls ($0.01/call) ↗
MCP tools: 2: resolve-library-id, query-docs ↗		Maintained by Upstash; repo https://github.com/upstash/context7. Local server @upstash/context7-mcp, also a ctx7 CLI and @upstash/context7-sdk. REST base URL: https://context7.com/api/. Exceeding the limit returns 429 with Retry-After/RateLimit-* headers. Paid plans cover teams and private repos.Smithery 5,563 uses
Crossref REST APILook up and search scholarly metadata such as DOIs, references, authors, funders and licenses across Crossref-registered works.none
no key; add mailto=<email> param or User-Agent contact for the 'polite pool'		communityfree, no keyBase URL: https://api.crossref.org
Free plan: no_key, no card ↗
Rate limits: Public pool 5 req/s, concurrency 1; polite pool (mailto) 10 req/s, concurrency 3; Metadata Plus 150 req/s ↗
Example call: search scholarly works metadata: GET https://api.crossref.org/works?query.bibliographic={query}&rows={n}&mailto={email} ↗
MCP tools: 3: search_papers, fetch_paper_details, search_by_topic ↗		Base URL https://api.crossref.org (/works, /journals, /funders, /members). Rate limits are fixed per pool: public 5 req/s (concurrency 1), polite with mailto 10 req/s (concurrency 3); 429 when exceeded. Metadata Plus is a paid SLA tier (150 req/s).GitHub 117★ afrise/academic-search-mcp-server
GDELT APIFull-text search global news in about 65 languages and analyze coverage volume, tone and geography in near real time.none
no auth		communityfree, no keyBase URL: https://api.gdeltproject.org/api/v2/doc/doc
Free plan: no_key, no card ↗
Example call: search recent global news articles: GET https://api.gdeltproject.org/api/v2/doc/doc?query={query}&mode=artlist&maxrecords={n}&format=json ↗
MCP tools: 9: gdelt_search_articles, gdelt_get_coverage_timeline, gdelt_get_tone_distribution, gdelt_get_coverage_breakdown, gdelt_search_tv, gdelt_get_tv_clips … ↗		DOC 2.0 API: https://api.gdeltproject.org/api/v2/doc/doc. Covers a rolling window of about 3 months, max 250 records per query. GEO 2.0 and TV APIs also exist. Full historical Events/GKG data available via Google BigQuery (needs a billed Google Cloud account).GitHub 6★ cyanheads/gdelt-mcp-server
Gitingest APILets an agent turn any public Git repository into a single text digest (summary, directory tree and file contents) sized for an LLM prompt.none
No key for public repos: GET https://gitingest.com/api/<owner>/<repo>; optional GitHub token for private repos via the 'token' parameter		communityNo key needed: free public service; hosted API limited to 10 requests/minute per endpoint (the web UI also has rate limiting and CAPTCHA protection).Base URL: https://gitingest.com/api
Free plan: no_key, no card ↗
Rate limits: 10 requests/minute on POST /api/ingest and GET /api/{user}/{repository} (per server limiter) ↗
Example call: turn a public GitHub repo into a text digest: GET https://gitingest.com/api/{user}/{repository} ↗
MCP tools: 1: ingest_git ↗		GET /api/{user}/{repo} or POST /api/ingest (include/exclude patterns); digest returned inline (summary, tree, content) plus a download URL. For heavy use the vendor recommends the open-source CLI (pip install gitingest).—
Hacker News APIRead Hacker News stories, comments, jobs, polls, user profiles and front-page rankings as real-time JSON.none
no auth		communityfree, no keyBase URL: https://hacker-news.firebaseio.com/v0/
Free plan: no_key, no card ↗
Rate limits: There is currently no rate limit ↗
Example call: get a story or comment by id: GET https://hacker-news.firebaseio.com/v0/item/{id}.json ↗
MCP tools: 1: get_stories ↗		Base URL https://hacker-news.firebaseio.com/v0/. Read-only, no documented rate limit. No full-text search; use the separate Algolia HN Search API for that.GitHub 13.3k★ HackerNews/API
GitHub 40★ pskill9/hn-server
Internet Archive Open Library APILets an agent search books, authors and subjects, fetch work/edition data by ISBN or OLID, search inside book texts and get cover images.none
No key; send a User-Agent header with app name and contact email, e.g. 'User-Agent: MyApp (me@example.org)'		communityNo key; free and open: 1 request/second anonymous, 3 requests/second with an identifying User-Agent.Base URL: https://openlibrary.org
Free plan: no_key, no card ↗
Rate limits: 1 request/second unidentified; 3 requests/second with User-Agent naming app and contact; no bulk download via API ↗
Example call: search books: GET https://openlibrary.org/search.json?q={query} ↗
MCP tools: 7: search_books, get_book_by_title, get_authors_by_name, get_author_info, get_author_photo, get_book_cover … ↗		Bulk scraping discouraged; use the monthly data dumps for large jobs. Reading-log/list write endpoints need an Open Library login.—
New York Times APIsLets an agent search NYT articles back to 1851, read top stories, most-popular lists, the Times Wire feed and bestseller lists.API key
Query parameter ?api-key=<API_KEY> (create an app at developer.nytimes.com and enable the APIs you need)		communityFree: all public APIs are free; 500 requests/day and 5 requests/minute per API.Base URL: https://api.nytimes.com/svc
Free plan: 500 requests per API, per day, no card ↗
Rate limits: Two limits per API: 500 requests/day and 5 requests/minute; sleep 12 s between calls ↗
Data policy: Non-commercial use only; terms forbid using the APIs or content to develop a software product or train an ML/AI system ↗
Example call: search articles: GET https://api.nytimes.com/svc/search/v2/articlesearch.json?q={query}&api-key={API_KEY} ↗
MCP tools: 5: search_articles, get_latest_news, get_most_popular, get_archive, get_bestseller_list ↗		Returns metadata/abstracts and links, not full article text. Exceeding limits returns HTTP 429; wait ~12 s between calls. Non-commercial use only; terms forbid using the APIs to develop a software product or train an ML/AI system.—
NewsAPISearch and retrieve current and historical news articles and top headlines from thousands of sources worldwide.API key
X-Api-Key: <key> header or apiKey=<key> query param		communityfree Developer plan: 100 req/day, dev/test only, 24h article delayBase URL: https://newsapi.org/v2
Free plan: 100 requests, per day, no card ↗
Rate limits: Developer plan 100 requests/day, 24h article delay, dev/test only; Business $449/mo with 250,000 requests/month ↗
Example call: search news articles: GET https://newsapi.org/v2/everything?q={query} ↗
Unit price: Extra requests beyond plan: $0.0018 per request on Business ($449/mo incl. 250,000), $0.0009 on Advanced ($1,749/mo incl. 2M); free Developer plan has no overage ↗
MCP tools: 2: search_articles, get_top_headlines ↗		Free plan is barred from staging/production, including internal use; CORS allowed from localhost only. Production use needs the Business plan ($449/mo, 250k req/mo). Endpoints: /v2/everything, /v2/top-headlines.GitHub 3★ matteoantoci/mcp-newsapi
OpenAlex APIQuery an open catalog of scholarly works, authors, institutions, sources and topics, including citation data, via REST.API key
api_key=<key> query param (optional; basic use works without key)		communityfree: $1 of API usage/day without payment method; free key raises daily budget 10xBase URL: https://api.openalex.org
Free plan: $1 of API usage (10x with free API key), per day, no card ↗
Rate limits: Cost-based daily budget ($1/day free, resets midnight UTC); each response reports cost_usd; prepaid top-ups in $1 increments ↗
Data policy: Data released under CC0 ↗
Example call: search scholarly works: GET https://api.openalex.org/works?search={query} ↗
Unit price: search: $1 per 1,000 calls; list+filter: $0.10 per 1,000; single entity: free; content download: $10 per 1,000 (after $1/day free usage) ↗
MCP tools: 5: list_categories, fetch_latest, fetch_top_cited, search_papers, fetch_content ↗		Base URL https://api.openalex.org. docs.openalex.org redirects to help.openalex.org. Free daily budget resets at midnight UTC; then pay-as-you-go in prepaid $1 increments. Member plans from $5,000/yr. Data is CC0. Listed MCP is a community multi-source papers server (arXiv, OpenAlex, PMC and more).GitHub 57★ benedict2310/Scientific-Papers-MCP
PubMed E-utilities (NCBI)Search, fetch, summarize and cross-link biomedical literature records from PubMed, PMC and other Entrez databases.none
optional api_key=<NCBI key> query param; tool and email params recommended		communityfree; 3 req/s without key, 10 req/s with free NCBI API keyBase URL: https://eutils.ncbi.nlm.nih.gov/entrez/eutils/
Free plan: no_key, no card ↗
Rate limits: 3 requests/second per IP without key; 10 requests/second with free API key (one key per NCBI account) ↗
Example call: search PubMed for article IDs (ESearch): GET https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=pubmed&term={query}&retmode=json ↗
MCP tools: 5: search_papers, search_pubmed, search_pmc, search_europe_pmc, download_with_fallback ↗		Base URL https://eutils.ncbi.nlm.nih.gov/entrez/eutils/ (esearch, efetch, esummary, elink). Free API key from NCBI account settings. Run heavy jobs off-peak.GitHub 2,692★ openags/paper-search-mcp
Semantic Scholar APISearch papers, authors and citation graphs, and get paper recommendations from Semantic Scholar's academic corpus.API key
x-api-key: <key> (optional; unauthenticated access allowed)		communityfree; no key (shared, throttled pool) or free API key on requestBase URL: https://api.semanticscholar.org/graph/v1
Free plan: no_key, no card ↗
Rate limits: Unauthenticated: 1000 req/s shared among all users, may be throttled; with key: 1 req/s on all endpoints to start ↗
Example call: search papers by relevance: GET https://api.semanticscholar.org/graph/v1/paper/search?query={query}&fields={fields} ↗
MCP tools: 16: paper_relevance_search, paper_bulk_search, paper_title_search, paper_details, paper_batch_details, paper_authors … ↗		Without a key, all users share a pool of 1000 req/s that can be throttled under load. A key starts at 1 req/s dedicated; requested via web form, arrives by email (not instant). Three APIs: Academic Graph, Recommendations, Datasets.GitHub 165★ zongmin-yu/semantic-scholar-fastmcp-mcp-server
WikidataQuery the Wikidata knowledge graph of entities, facts and relations via SPARQL, the REST API or MCP.none
no auth for reads (descriptive User-Agent required); OAuth2 Bearer only for edits		officialfree, no keyBase URL: https://query.wikidata.org/sparql
Free plan: no_key, no card ↗
Rate limits: Wikimedia limits: 10 req/min IP-only, 200 req/min with proper User-Agent; back off on 429 and honor Retry-After ↗
Example call: run a SPARQL query: GET https://query.wikidata.org/sparql?query={sparql} ↗
MCP tools: 6: search_items, search_properties, get_statements, get_statement_values, get_instance_and_subclass_hierarchy, execute_sparql ↗		Official MCP comes from Wikimedia Deutschland (github.com/wmde/WikidataMCP). SPARQL endpoint: https://query.wikidata.org/sparql, query timeout ~60s. REST API at /w/rest.php/wikibase/v1. Wikimedia rate limits and User-Agent policy apply. Added because it's a widely-used structured-knowledge source missing from the candidate list.—
Wikipedia / Wikimedia APISearch, read and summarize Wikipedia and other Wikimedia content and metadata via REST and Action APIs.none
Reads need no key but require a descriptive User-Agent header with contact info. Optional OAuth2 via Authorization: Bearer <access token> raises rate limits.		communityfree, no key (rate-limited)Base URL: https://en.wikipedia.org/w/api.php
Free plan: no_key, no card ↗
Rate limits: 10 req/min IP-only; 200 req/min with User-Agent (unauthenticated or new users); 2000 req/min established editors; max 3 concurrent ↗
Data policy: Text available under CC BY-SA; additional terms may apply ↗
Example call: full-text search Wikipedia articles: GET https://en.wikipedia.org/w/api.php?action=query&list=search&srsearch={query}&format=json ↗
MCP tools: 9: search_wikipedia, get_article, get_summary, get_sections, get_links, get_coordinates … ↗		The old API Portal at api.wikimedia.org now 301-redirects to mediawiki.org/wiki/Wikimedia_APIs. Documented limits: 10 req/min for requests identified only by IP, 200 req/min for unauthenticated requests with a proper User-Agent, 2000 req/min for established editors. Main endpoints are the per-wiki MediaWiki Action API (/w/api.php) and REST APIs.GitHub 295★ Rudra-ravi/wikipedia-mcp
Wolfram|Alpha APIGet computed answers, structured data and step-by-step results for math, science and factual queries.API key
appid=<AppID> query param or Authorization: Bearer <AppID>		officialFree: non-commercial developer AppID with up to 2,000 calls/month; paid for commercial useBase URL: https://www.wolframalpha.com/api/v1/llm-api
Free plan: 2,000 non-commercial API calls, per month ↗
Example call: get a computed answer for a natural-language query (LLM API): GET https://www.wolframalpha.com/api/v1/llm-api?input={query}&appid={APPID} ↗
MCP tools: 3: WolframAlpha, WolframLanguageEvaluator, WolframContext ↗		LLM API https://www.wolframalpha.com/api/v1/llm-api is built for agents (maxchars, default 6800); get an AppID in the Developer Portal; commercial use needs a paid license. Official MCP: free, no auth at agenttools.wolfram.com/mcp; paid services.wolfram.com/api/mcp (Bearer).—

Other categories