papersgpt/papersgpt-for-zotero

What it solves

PapersGPT is designed for researchers, legal professionals, and finance experts who need to synthesize information across thousands of PDFs in their Zotero library without compromising data privacy. It solves the problem of slow, imprecise "fuzzy" semantic search and the risk of data leakage associated with cloud-based AI tools.

How it works

The tool is a Zotero extension with a high-performance C++ core that performs parsing, indexing, and searching locally on the user's machine. It enables users to chat with a single PDF, a selection of documents, or their entire library. It supports local LLMs (via integrated models, Ollama, or custom OpenAI-compatible endpoints) to ensure a fully air-gapped experience, and provides interactive citations that link directly to the source text in the PDF.

Who it’s for

Academic researchers, legal professionals, and finance analysts who manage large-scale document libraries and handle sensitive data that requires absolute data sovereignty.

Highlights

• Industrial-Scale Performance: Uses a C++ core for near-instant indexing of 10,000+ documents with minimal RAM usage.
• Zero-Byte Data Leakage: Supports 100% offline local LLMs for complete privacy.
• AutoPilot Automation: An autonomous researcher agent that can batch-process thousands of papers overnight and save insights into Zotero Notes.
• Zotero MCP Server: Allows integration of the Zotero knowledge base as a skill for agentic platforms like Claude Code, Cursor, or Windsurf.
• Verifiable Evidence: Every AI response includes precise, clickable citations that jump directly to the specific paper.