strato-mcp-zotero

A Python MCP server that connects your Zotero library to AI assistants. Rather than exposing raw API endpoints, it offers a small set of tools shaped around how researchers actually work: finding sources, browsing collections, inspecting items, and saving or updating metadata.

Works with Claude, Gemini CLI, OpenAI Codex, and any MCP-compatible client.

This server is built around how researchers use Zotero, not around the API surface. Instead of exposing raw endpoints, it offers five tools shaped by real workflows:

1. Find sources - find_library_sources searches your library by keyword, DOI, or author and returns compact summaries with optional collection context. The natural starting point for any research question.
2. Inspect a source - Once you have an item key, inspect_saved_source returns its full normalized metadata. Useful for checking what's already saved before adding a duplicate.
3. Browse a collection - review_collection resolves a collection by name or key, summarizes it, and lists its top items (plus child collections if you ask). Good for "what do I have on X?"
4. Save a source - save_source_to_library takes a title, creators, tags, and collection assignments and adds the item to your library, returning a summary of what was saved.
5. Update metadata - update_saved_source patches only the fields you provide: fix a DOI, add tags, move to a different collection, without touching anything else.

Collection arguments accept either a name or a key. If a name is ambiguous, the tool returns candidates so the agent can resolve it.

For save_source_to_library and update_saved_source, use ; to separate creators and , for tags:

creators="Ada Lovelace; Grace Hopper"
creators="author: Turing, Alan; editor: Knuth, Donald"
tags="reading-queue, llm, bibliography"

Zotero API key:

1. Go to your Zotero account settings → API Keys
2. Create a key with