Search research, books and discussion with citable sources
Full-text retrieval across papers, books, patents, Wikipedia, and public discussion — sourced answers with canonical citations, for people and AI agents
Answers you can check
Every answer cites canonical source URIs — stable links to the documents, so you can open them and verify. Retrieved text is evidence for review, not proof that a claim is correct.
Sources in Claude and ChatGPT →
Connect the assistant you already use so its answers cite papers, books, and patents you can open and check.
Citation-grounded agents →
Give an agent compact search results, canonical source URIs, bounded full text, and document-level passage search.
Systematic reviews →
Build a traceable discovery and screening pipeline across scholarly metadata, abstracts, and available full text.
Patent prior-art search →
Search patents alongside papers and technical reference works, then preserve identifiers for expert review.
Engineering standards →
Find ISO, British, DIN and GOST standards by number, part and edition year, then read their text through the API or MCP.
Build on the corpus
The same retrieval in three ways: the web app to prototype, REST to automate, MCP for agents. One key, per-request pricing.
Search API
Ground your agents and models with the same retrieval over REST or MCP.
import requests
resp = requests.post(
"https://api.machinelibrary.ai/v2/search/",
json={
"query": "breast cancer first-line drugs",
"limit": 10,
},
headers={"Authorization": f"Bearer {API_KEY}"},
)
for doc in resp.json()["hits"]:
print(doc["document"]["title"])The corpus, in bulk
multi-billion documents across papers, books, patents, encyclopedias, and public discussion — collected, parsed, and embedded. Licensable for research and AI training.
Scholarly papers
Journal articles, proceedings and preprints, with full text where available
Books
Millions of publicly available books in various formats
Patents
Patent publications with their stored text
Encyclopedias
Wikipedia and other encyclopedia and reference entries
Magazines
Magazine and periodical archives
YouTube
Transcripts and metadata from public video content
Public discussions and community threads
Telegram
Messages and media from public channels
Dataset licensing
Bulk access is licensed per corpus and use — research, AI training, or commercial applications. See what is inside first; the methodology and figures are public.