Labsco
MCP SERVER

Search, read and cite 22,000+ rare pre-modern books with English translations — with exact page quotes and citation URLs.

Encyclopedic & Cultural ReferenceVerified
Summary
Quoting is its own tool, which is exactly the right call for historical sources.

get_quote returns one page's verbatim translation with the original OCR and a formatted citation, so an assistant stops reconstructing quotations from memory — the failure that makes AI-assisted historical research untrustworthy. Around it the searches divide cleanly: metadata for what exists, translated text for a phrase, and concept search for ideas these authors expressed in words nobody uses today.

What it is

A research server over the Source Library collection: rare pre-modern texts on alchemy, Hermetica, Kabbalah, theology and early science, with AI-generated English translations alongside the original OCR, plus a large image collection.

What you get
  • Full-text search across titles, authors, translations and OCR, returning books and page snippets with citation URLs — search_library
  • The catalog browsed by author or title fragment, language, category or translation recency, with page counts and translation progress — list_books
  • Search inside the translated text of the whole library, for sweeping one theme across many books — search_translations
  • Conceptual search for when the modern term never appears literally — a query about distributed cognition reaching passages on active intellect and the art of memory — search_concept
  • Search confined to one book's pages, across both OCR and translation — search_within_book
  • A book's summary, chapter list, edition metadata, DOI, page counts and processing status — get_book
  • The text itself, read a chapter or a page range at a time, with page markers for citation and an explicit flag when the response was truncated — get_book_text
  • The verbatim translation of a single page, with the original OCR and a formatted citation, for quoting — get_quote
  • A visual collection searched by type, subject, figure, symbol, year or book: 50,000+ illustrations lifted from pages plus 23,000+ standalone artworks from the Met, the Rijksmuseum, Wikimedia and the NGA — search_images
  • A duplicate check before importing a book, with a confidence level and a recommendation — check_duplicate
  • Research shared back as a dossier of references, and feedback sent to the maintainers — share_findings, submit_feedback
Requirements

No API key. Run npx -y @source-library/mcp-server over stdio, or install the package globally — the same install doubles as a CLI, so a search can also be run straight from a terminal. Building from source is a clone, npm install and npm run build.

Setup effort

One command — npx -y @source-library/mcp-server