Lectern

Connect an AI

MCP over HTTP4 toolsRead-onlyNo key

Claude Code
claude mcp add --transport http lectern https://lectern-api.restorationcommons.org/mcp
Claude.ai, ChatGPT, others
https://lectern-api.restorationcommons.org/mcp

Tools

  • search(query, collection?, org?, date_from?, date_to?, doc?, mode?, sort?, limit?)

    Ranked pages. Words are ANDed; "a phrase", prefix*, a NEAR/5 b and -exclude work. Filter by shelf (collection), publisher, year or one book. Every hit is a citation.

  • fetch(id)

    A book id returns its record and contents. A cite string returns the page, in full, with a link to its image.

  • verify_quote(quote, cite?, doc?)

    Whether the words are at the cited place. Five verdicts: exact and normalized pass; elsewhere, near_miss and not_found fail, with where the words really are or what the page says instead.

  • check_answer(text)

    Audits a draft answer: every citation in it, the quotations attached to each, and how each one checked out.

A citation

npnf2:04-select-works-and-lettersbook/r1scan/p65page of the scan
https://lectern.restorationcommons.org/d/npnf2/04-select-works-and-letters/r1/p65

HTTP

GET/api/search?query=…SearchResponse
GET/api/doc/<ns>/<id>[/<version>]DocumentCard
GET/api/units/<ns>/<id>/<version>/<anchor>UnitText[]
POST/api/verifyQuoteCheck
POST/api/checkAnswerAudit

JSON SchemasThe engine's architecture