List every Zotero library this server can see: your personal library ("My Library", library_id 1) plus every group/shared library you belong to, with its library_id, name, and type. Pass a group's library_id as the library_id argument to list_collections, list_items, search_metadata, get_tags, search_library, or get_library_stats to scope to that shared library specifically — otherwise those tools already search across every library by default. Item-level tools (get_item, get_item_fulltext, get_item_notes, get_item_annotations, compare_items) work the same regardless of which library an item lives in — just pass its key.

List every collection (and sub-collection) in the Zotero library as a flat list with parent links, so Claude can build or walk the collection tree. Each entry has key, name, library_id, and parent_key (null for top-level collections). Defaults to every library (your personal library plus all group/shared libraries); pass library_id (see list_libraries) to scope to just one.

List items (papers, books, etc. — not notes/attachments) in the library, most recently modified first. Filter by collection_key (from list_collections), tag (exact tag name), and/or item_type (e.g. 'journalArticle', 'book', 'conferencePaper'). Use limit/offset to page through a large library. Defaults to every library (personal plus all group/shared libraries); pass library_id (see list_libraries) to scope to just one. Returns key, title, library_id, item_type, creators_summary, year, dateAdded, dateModified for each item.

Get full metadata for one item: all bibliographic fields, creators, tags, the collections it belongs to, its abstract, and a summary of its child notes and attachments (so you know what's available to fetch next with get_item_fulltext / get_item_notes / get_item_annotations). Works for items in any library — personal or group/shared — the response's library_id says which one this item lives in.

Fast search over item titles, abstracts, and creator names (does NOT search inside PDFs or notes — use search_library or search_fulltext for that). Good for quickly locating an item you already know roughly. Searches your personal library and every group/shared library by default; pass library_id (see list_libraries) to scope to one.

Extract the full text of an item's PDF attachment(s), with [p. N] page markers so you can cite specific pages back to the user. Optionally restrict to a page range with page_start/page_end (1-indexed, inclusive) — useful for a long document where you only need a section. max_chars caps the response size (defaults to a safe per-item limit); the result says so and suggests narrowing the page range if truncated. This extracts on demand — it does not require the search index to have been built first.

Get all notes attached to an item (child notes), or the content of a standalone note if item_key is itself a note. Note content is converted from Zotero's internal HTML to plain text.

Get every PDF annotation (highlights, notes, underlines) the user has made on an item's PDF: the highlighted/underlined text, any comment the user added, the highlight color, and the page label — in reading order. This surfaces what the user found important, which is often more useful for synthesis than the raw PDF text.

List tags. If item_key is given, list that item's own tags. If omitted, list every tag used anywhere in the library with how many items carry it, most-used first — handy for discovering the user's own taxonomy before filtering list_items/search_library by tag. When item_key is omitted, defaults to aggregating across every library (personal plus all group/shared libraries); pass library_id (see list_libraries) to scope to just one.

Search across the whole library. scope is one of: 'metadata' (titles/authors/abstracts, always available), 'fulltext' (inside PDF text, requires the index to have been built — see rebuild_search_index / scripts/build_index.py), 'notes' (inside your notes, same index requirement), or 'all' (default: every scope, merged). Searches across every library — your personal library and all group/shared libraries — by default; pass library_id (see list_libraries) to scope to just one. Each fulltext/notes result includes a library_id, a snippet, and the exact page (for PDFs) so you can cite it, plus the parent item_key to fetch more with get_item / get_item_fulltext. IMPORTANT: whenever you report fulltext or notes results to the user, mention the index_last_updated timestamp included in the response (e.g. "based on your library as of ") — those two scopes come from a separately built search index, not a live query, so the user should know how current it is. metadata-only results don't need this since they're always live.

(Re)build the full-text search index used by search_library's 'fulltext' and 'notes' scopes. Indexes every library — personal plus every group/shared library — by default, so shared libraries become full-text searchable automatically; pass library_id (see list_libraries) to index just one. Incremental by default (full=False): only new or changed PDFs/notes since the last run are processed, which is fast. Pass full=True to wipe and rebuild everything from scratch. For a large library, prefer running python scripts/build_index.py in a terminal the first time, since that gives you progress output; this tool is best for small incremental top-ups from within a conversation.

Fetch metadata (and optionally truncated full text) for several items in one call, so you can compare/synthesize across them — arguments, methods, findings, etc. Items can come from different libraries (personal and/or group/shared) in the same call — each entry's library_id says which one it's from. Set include_fulltext=True to also pull each item's PDF text (capped at max_chars_per_item per item to keep the combined response manageable); leave it False to compare on metadata, abstracts, tags and annotations only.

Overall library stats: item/collection/tag counts and full-text index coverage (how many PDFs/notes are searchable via search_library's fulltext/notes scopes right now), including fulltext_index.index_last_updated. Omit library_id to get combined totals across your personal library and every group/shared library, with a by_library breakdown (see list_libraries for names); pass a specific library_id to scope to just one. Mention the index_last_updated timestamp to the user whenever you report these stats or otherwise rely on the full-text index.