papersgpt/papersgpt-for-zotero

⭐ 2,349  ·  JavaScript  ·  GitHub Repo

A powerful Zotero AI and MCP plugin with ChatGPT, Gemini 3.1, Claude, DeepSeek V4, Grok, OpenRouter, Kimi 2.5, GLM 5, SiliconFlow, GPT-oss, Gemma 4, Qwen 3.5

ai chat chatgpt claude deepresearch deepseek gemini gemma4

1-Sentence Summary

Zotero plugin turning your literature library into an autonomous, multi-model AI research assistant with offline capabilities.

🔥 Key Capabilities & USP

• AutoPilot Autonomous Agent: Select dozens of papers, define a research goal, and let the AI work overnight—reading, analyzing, and writing structured notes directly into Zotero. Solves the pain point of manual literature review at scale.
• Multi-LLM & Local Model Support: Access Gemini 3.1, GPT 5.5, Claude Opus 4.6, DeepSeek V4, Grok 4, Qwen 3.5 via cloud APIs, or run Gemma 4, Phi-4, GPT-oss-20B entirely offline with one-click download. Solves data privacy concerns and API cost issues.
• MCP (Model Context Protocol) Integration: Any MCP-compatible chatbot (Claude, ChatGPT desktop) can connect directly to your personal Zotero library. Solves the pain point of AI tools being disconnected from your research database.
• Multi-PDF Chat with Blazing-Fast Performance: Chat with single or multiple 100+ page PDFs simultaneously, with 5x faster reading speed. Solves the pain point of slow document analysis in traditional tools.
• Cross-Platform & Zotero 8/9 Compatible: Works on Windows, Mac, and Linux, supporting the latest Zotero versions. USP: No other Zotero plugin combines autonomous batch processing, local LLM execution, and MCP integration in a single package.

Technical Architecture

Component	Details
Plugin Type	Zotero XPI plugin (JavaScript/Node.js)
LLM Backend	Cloud APIs (OpenRouter, SiliconFlow, gpt-oss, Kimi, Z.ai) + Local models (HuggingFace/GitHub downloads)
Local RAG Stack	On-device embeddings, vector database, and rerank modules for 100% privacy
Model Execution	One-click model download with automatic GPU detection; Ollama compatible
Build System	npm install && npm run build → produces papersgpt.xpi
External Integration	MCP protocol for chatbot clients to query Zotero library

Quick Start Guide

# Clone the repository
git clone https://github.com/papersgpt/papersgpt-for-zotero.git

# Navigate to project directory
cd papersgpt-for-zotero

# Install dependencies
npm install

# Build the plugin
npm run build

After building, install the generated papersgpt.xpi file into Zotero (Tools → Add-ons → Install Add-on From File). Configure your preferred LLM API keys or download a local model from the plugin settings.

Pros, Cons & Use Cases

Pros

• Unmatched flexibility: Supports both cloud and local LLMs, giving users control over cost, privacy, and model choice.
• True autonomy: AutoPilot batch processes hundreds of papers without manual intervention—a game-changer for literature reviews.
• Seamless Zotero integration: Notes are saved directly into your Zotero library, not a separate app.
• MCP ecosystem: Extends AI capabilities beyond the plugin to any MCP-compatible client.
• Active development: Supports Zotero 8 and 9, with frequent updates.

Cons

• AutoPilot in open beta: Daily usage limits apply during the beta phase.
• Local model downloads: Require stable internet connection to HuggingFace/GitHub; large model files may take time.
• Advanced model access: Some models (e.g., GPT 5) require identity verification.
• Learning curve: Users unfamiliar with LLM APIs or local model setup may need initial guidance.

Who should NOT use this?

• Users who never read PDFs in Zotero: If you only use Zotero for citation management without reading papers, the AI features are overkill.
• Privacy-paramoid users unwilling to use local models: Cloud API calls send paper text to third-party servers unless you exclusively use local models.
• Researchers with zero interest in AI/LLMs: The plugin's value is entirely AI-driven; manual-only users won't benefit.
• Users on unsupported Zotero versions: Requires Zotero 8 or 9; older versions are incompatible.

Ideal Use Cases

• Literature review marathons: PhD students or researchers reviewing 50-100 papers for a thesis or grant proposal.
• Systematic reviews: Teams needing consistent, structured summaries across hundreds of papers.
• Privacy-sensitive research: Medical, legal, or classified research where data cannot leave the local machine (using local models).
• Multi-model experimentation: Researchers comparing outputs from Gemini, Claude, DeepSeek, and local models for the same papers.
• AI-assisted reading workflows: Users who want to chat with their entire Zotero library through MCP-compatible chatbots.

Community & Activity

With 2,349 stars and active maintenance (last updated May 2026), PapersGPT has strong momentum in the academic AI community. The project's rapid feature iteration—adding support for the latest models like Gemma 4, Qwen 3.5, and GPT-oss—signals a responsive development team. The growing ecosystem of MCP integration and local model support positions it as a leading tool for AI-powered research management. If you're in the Zotero ecosystem, this is a project worth watching (and starring).