# hebrewbooks.ai

> Morphological full-text search over ~50,000 OCR'd Hebrew books from the
> HebrewBooks.org archive. Hebrew queries are matched by lemma (DictaBERT) so
> prefixes (ב/כ/ל/מ/ה/ו/ש) and inflections are handled automatically; non-Hebrew
> queries (English, Russian, Spanish, French, …) are auto-translated to Hebrew.
> Results collapse to books, each with matched-page snippets that deep-link to the
> source page on HebrewBooks.org.

## API (JSON)

- [OpenAPI spec](https://hebrewbooks.ai/api/openapi.json): machine-readable description of every endpoint
- [Interactive docs (Swagger)](https://hebrewbooks.ai/api/docs)
- Search: `GET https://hebrewbooks.ai/api/search?q={query}&scope={all|title|author}&size={n}`
  → JSON: `{ total_books, total_pages, results: [ { book_id, title_he, author_he, year, topic, pages: [ { pgnum, source_url, snippet } ] } ], suggestions }`.
  Add `translate=false` to skip auto-translation (and its LLM call).
- Filters (compose with any search): `topic`, `author`, `year_from`, `year_to`, `tier`, `book`.
- Categories: `GET https://hebrewbooks.ai/api/facets` → book count per topic.

## For agents (MCP)

- MCP server (streamable HTTP): `https://hebrewbooks.ai/mcp`
- Tools: `search(query, scope, topic?, author?, year_from?, year_to?, limit?)`,
  `list_topics()`, `browse_topic(topic, limit?)`, `get_book(book_id)`.

## Notes

- Scopes: `all` (page body, highlighted snippets), `title`, `author`.
- The endpoint is rate-limited — please be considerate. Source page images © HebrewBooks.org.