Related Servers

Most tools have clearly distinct purposes (search vs. get vs. create vs. update vs. delete vs. synthesize), and the descriptions are detailed enough to disambiguate similar operations like zotero_search_items vs. zotero_semantic_search vs. zotero_advanced_search. However, the sheer number of search/retrieval variants (search_items, search_by_tag, advanced_search, semantic_search, search_collections, search_by_citation_key) creates some boundary confusion, and zotero_get_item_metadata vs. zotero_get_item_fulltext vs. zotero_get_notes vs. zotero_get_annotations require careful reading to pick correctly.

All tools follow a consistent zotero_verb_noun pattern with snake_case throughout. Verbs are predictable (get, search, create, update, delete, add, set, manage, synthesize, export, switch, list, attach, read, batch_update), and nouns map to Zotero domain objects (item, collection, annotation, note, attachment, library, tag). This is a model of consistent naming.

38 tools is on the heavy side for a single MCP server, though Zotero is a complex domain with many entity types (items, collections, notes, annotations, attachments, libraries, tags, PDFs, semantic search, bibliography export). The count is justified by the breadth, but it exceeds the typical well-scoped range and will require an agent to navigate a large tool surface.

The tool surface covers the full lifecycle for Zotero's core entities: items (add, get, update, delete, search, batch update), collections (create, delete, list, search, get items), notes (get, manage), annotations (create, get, update, delete, synthesize), attachments (attach, get path, read pages, get outline), libraries (list, switch), plus advanced capabilities like semantic search, PDF layout detection, and bibliography export. There are no obvious dead ends; every read operation has a corresponding write operation where appropriate.