Agntic AI for Research Papers

MCP Agntic AI for Research Papers

This project implements a chatbot using the Model Context Protocol (MCP) to search and retrieve information about research papers from arXiv. The chatbot allows you to query papers by topic and extract detailed information about specific papers.

Overview

The system consists of two main components:

• Server: A FastMCP server that provides tools for searching arXiv papers and extracting paper information.

• Client: An MCP client that integrates with OpenAI's GPT model to process user queries and interact with the server.

The server stores paper information in JSON files organized by topic, while the client provides an interactive chat interface for users to input queries.

Features

• Search Papers: Search for papers on arXiv by topic, with configurable maximum results.

• Extract Paper Info: Retrieve detailed information (title, authors, summary, PDF URL, publication date) for a specific paper using its arXiv ID.

• Persistent Storage: Paper information is saved in JSON files under a papers directory, organized by topic.

• Interactive Chatbot: Users can interact with the chatbot via a command-line interface, with support for natural language queries powered by OpenAI's GPT model.

Project Structure

├── papers/ # Directory for storing paper information (auto-created)
├── src/
│ ├── mcp_chatbot.py # MCP client with chatbot implementation
│ ├── research_server.py # FastMCP server with arXiv search tools
│ ├── keys.json # API keys (not tracked in git)
│ ├── server_config.json # MCP server configuration
├── README.md
├── main.py # Entry point

Example Queries

• Search for papers:

Query: Find 3 papers on machine learning 

Output: List of paper IDs, with details saved in papers/machine_learning/papers_info.json.

• Extract paper information:

Query: Get info for paper 2103.12345 

Output: JSON-formatted paper details (title, authors, summary, etc.) if found.

Notes

• The server creates a papers directory to store JSON files containing paper information, organized by topic (e.g., papers/quantum_computing/papers_info.json).

• The client uses gpt-4o-mini by default. Update the model in src/mcp_chatbot.py if needed.

• The system assumes uv is installed for running scripts. Modify the command in server_config.json if using a different tool (e.g., python).

Future Improvements

• Add support for filtering papers by date, author, or category.

• Implement paper PDF download and storage.

• Enhance the chatbot with more natural language understanding for complex queries.

• Add a web-based UI for better user interaction.

Requirements

• Python 3.12+

• Dependencies (install via uv or pip):

• arxiv

• mcp

• openai

• nest-asyncio

• python-dotenv

• OpenAI API key (stored in src/keys.json)

• uv (recommended, for running the server and client)

Installation

• Clone the repository: git clone cd

• Install dependencies using uv (recommended): uv pip install -r pyproject.toml Or with pip: pip install -r pyproject.toml

• Create a src/keys.json file with your OpenAI API key: { "open_ai_api": "your-openai-api-key" }

• Ensure the MCP server configuration in src/server_config.json is set up correctly: { "mcpServers": { "filesystem": { "command": "npx", "args": [ "-y", "@modelcontextprotocol/server-filesystem", "." ] }, "research": { "command": "uv", "args": ["run", "research_server.py"] }, "fetch": { "command": "uvx", "args": ["mcp-server-fetch"] } } }

Usage

• Start the MCP server: uv run src/research_server.py This runs the server with the research configuration, providing tools for paper search and extraction.

• Run the client in a separate terminal: uv run main.py The client connects to the server, initializes the chatbot, and starts the interactive chat loop.

• Interact with the chatbot:

• Enter a query like "Search for papers on quantum computing" or "Get info for paper 1234.56789".

• Type 'quit' to exit.

License

This project is licensed under the MIT License. See the LICENSE file for details.