For researchers using AI assistants
Real, checkable sources inside Claude and ChatGPT
Connect Machine Library to the assistant you already use. It searches papers, books, patents, standards, and Wikipedia instead of answering from memory, and links each claim to a source you can open.
A practical retrieval workflow
01
Connect once
Add Machine Library as a connector in Claude or ChatGPT, or as an MCP server in Claude Code or Codex, and sign in with your account.
02
Ask in your own words
The assistant searches the corpus for your question and reads the most relevant documents before it answers.
03
Check the passages
Ask for the exact passage behind a claim; the assistant can search inside the cited document and quote it.
04
Keep the citations
Every result carries a DOI or canonical link you can open, cite, or add to your reference manager.
What the retrieval layer returns
- Papers, books, patents, standards, and Wikipedia in one search.
- Quoted passages from inside long documents.
- DOI and canonical source links for every result.
- Reddit, Telegram, and Discord discussion when you ask for it.
Important boundaries
- A cited passage supports your review; it does not prove the generated answer is correct.
- Coverage and full-text availability vary by source.
- Searches and document reads use account credit; see pricing.
- Check important claims at the source before relying on them.
Implementation paths
Start with the interface that matches your prototype. The source identifiers and document model stay consistent as you move from manual exploration to an agent or application.