io.github.caelum29/calibre-mcp

Internal endpoint for the in-chat cover-board widget. Returns the cached payload of a recent search. Do not call as the model — search results already contain everything this returns.

Build the semantic index for specific books (required: bookId, ids, or query — full-library indexing is deferred). Extracts, chunks, and embeds each book. Set keywordOnly=true (or when the embedding model is absent, it happens automatically) to build a keyword-only index that powers mode:"keyword" search with zero ML dependencies. Re-run after adding books; use force to re-index unchanged ones. Set prune=true to also drop index entries for books that no longer exist in the library (removals/merges leave searchable orphans behind).

Find probable duplicate books. mode=identical (exact title+authors) or similar (fuzzy) group the library (or ids/query subset) with a merge-safety score; mode=compare diffs 2+ ids field-by-field. Read-only — never merges.

Get full metadata (authors, ISBN, formats, comments, cover) for one book by its id or uuid (id and bookId are interchangeable — pass the bookId from a search result). Set include_cover=true only when the cover image itself must be inspected.

Extract a book's text as a capped, fenced excerpt; pass the returned nextCursor token verbatim to walk the whole book, or jump to a char position with offset (e.g. a search passage's charStart). Set structure=true for a chapter map with per-chapter cursors. To find text inside a book, use calibre_search scope=book.

List a book's figures — images the text references, identified by caption (Figure 1-2 / Рис. 3.1) — with page + caption per figure, then fetch chosen ones as images via indexes (≤3 per call). Captions let you pick figures before spending image tokens. In MCP Apps hosts the fetched figures also render for the user in an in-chat viewer, so you can point at them by figure number — but describe only what the pixels you received actually show.

Browse library categories (Authors, Tags, Series, Languages, Publisher). No field → list the categories; with a field → its values + counts, filterable via a valueFilter regex (case-insensitive by default; a leading inline flag like `(?i)` is accepted).

List the Calibre libraries available on the Content Server and which one is the default. Call this first to discover valid library names.

Internal endpoint for the in-chat widgets' Open button. Launches a book in the local Calibre viewer via the calibre:// scheme. Do not call as the model — use calibre_get_book to show a book instead.

Health check: confirms the MCP server can reach the running Calibre Content Server via calibredb, and reports semantic-search status (embedding model, dependency, index vector count). Returns library categories on success.

Scan books for metadata problems: missing fields, raw-filename titles, invalid ISBNs, author-sort mismatches, series gaps. Defaults to the whole library; narrow with ids/query or specific checks. Read-only.

Propose real metadata for a book with a missing/raw-filename title via online provider lookup (ISBN from its identifiers or text, else title/author) on Open Library and Google Books. Preview only — returns a changes object for calibre_update_book; never writes.

Find books by exact title, author, ISBN, tag, or Calibre query syntax (mode=meta), or by full text (mode=fts). scope=book returns short keyword snippets from inside one book; first hits often land in TOC/front matter — for definitional or topic questions within a book prefer calibre_semantic_search scope=book (ranked passages with char offsets). Use calibre_semantic_search for meaning/topic queries. For "how many" questions use countOnly=true (returns only the count).

Meaning-based search over the local index. scope=library ranks books; scope=book (needs bookId) ranks passages within one book. target=figures searches figure captions instead of text — use when the user asks for a diagram/chart/schema; hits point at calibre_get_figures for the pixels. mode=hybrid (default) fuses semantic + keyword matches; mode=vector is semantic-only; mode=keyword is exact keyword/FTS (no model at query time). All modes need an index built by calibre_build_index.