zotero-mcp-code

The Zotero MCP crashes when heavy data overwhelms the LLM context. You can prevent these crashes by executing Zotero queries inside a code execution environment to keep large datasets out of the context window.

To combine semantic and keyword searches without duplicates, use a search orchestrator that runs multi-strategy queries, automatically deduplicates the results, and ranks them by relevance before returning a concise list.

Yes, you can safely fetch 50 to 100 Zotero items by loading the large datasets within a code execution environment, which processes the heavy data and returns only a formatted, concise set to the LLM.

To retrieve and rank top research papers for literature reviews, execute a comprehensive search query that fetches items via multiple strategies, deduplicates the library entries, and ranks them to return the most relevant papers.

Zotero queries support combining semantic search, keyword matching, and tag-based filtering together. The results are automatically deduplicated and ranked to provide a structured Markdown list for your research.