MCP Walkthrough: From Basics to Full Chatbot

An MCP-powered research assistant demonstrating the Model Context Protocol (MCP) using FastMCP and OpenAI. This project acts as a step-by-step walkthrough, progressing from basic tool calling to a multi-server chatbot with resources, prompts, and Claude Desktop integration.

Progression

• Basic Tool Calling: OpenAI tool calling without MCP to understand the baseline.
• Single Server Chatbot: Replaces manual definitions with an MCP client that automatically discovers tools from the server.
• Multi-Server System: Connects to multiple MCP servers, automatically routing tool calls.
• Full-Featured Chatbot: Integrates MCP resources (like dynamic topics) and prompts alongside tools.

Features & Integrations

• Tools: Search arXiv papers and extract details.
• Resources & Prompts: List available folders, read dynamic resources, and generate structured research prompts.
• Claude Desktop: Integrated as a local and remote research server within Claude Desktop.
• Deployment: Features both stdio and Streamable HTTP transports, with remote deployment on Render.com.