The project embeds a large language model agent directly into the Zotero reader, enabling users to query their library, summarize academic papers, inspect figures, compare sources across papers, and save AI-generated notes. It supports standard API providers, local OpenAI-compatible models, WebChat, Codex App-Server, and Claude Code. The system uses retrieval-augmented generation and MCP (Model Context Protocol) to ground responses in the user’s Zotero library.