MyArxivDB

This repository provides a MCP server to organize personal research papers using an MCP (Modular Command Platform) server and semantic search. The platform enables efficient paper retrieval, automatic project assignment, and assistance in writing literature review sections using LLMs.

This server is mainly designed for Claude Desktop but may also work well with other MCP clients.

This project was carried out as part of the term project for the BKMS1 course(@SNU GSDS).

---

⚡️ Quickstart

The following quickstart guide is based on an Apple Silicon MacBook.

1. Install uv

curl -LsSf https://astral.sh/uv/install.sh | sh

2. clone the repository

git clone https://github.com/Jongbin-kr/MyArxivDB_MCP.git

3. Intsall dependencies & activate the virtual environment

uv sync
source /.venv/bin/activate

4. set up environment variables at .env file.

# .env
PINECONE_API_KEY = "YOUR PINECONE_API_KEY"
DB_NAME = "YOUR_DB_NAME"
DB_USER = "YOUR_DB_USERNAME"
DB_PASSWORD = "YOUR_DB_PASSWORD"
DB_HOST = "localhost"
DB_PORT = 4444

5. Intsall MCP server at Claude Desktop

   mcp install server.py
   

6. Done! The Claude desktop app will automatically detect the MCP server and you can start using i!

---

📌 Motivation

Researchers frequently accumulate large numbers of papers but lack tools to systematically organize them by topic or project. BKMS aims to:

• Automatically assign new papers to relevant projects using embeddings
• Allow semantic search for project-specific literature
• Assist in drafting sections like “Related Work” using LLMs

---

🛠️ Main functions

Our MCP server supports the following core capabilities:

• Crawling metadata and PDFs from arXiv using ID or URL using ArXiv API
• Embedding abstracts using Pinecone API(llama-text-embed-v2)
• Storing papers and projects in a PostgreSQL + pgvector DB
• Generating "Related Work" sections via LLM prompts

---

🎥 Workflow & Demo video

You can see our PPT and demo video in assets folder.

Brief overview of our project workflow and DB schema is as follows.

---

👨‍👩‍👧‍👦 Team

• 박연진
• 원종빈
• 정예준