SymptomAI-MCP: Symptom Checker with Model Context Protocol

Overview

This project demonstrates an end-to-end architecture using the Model Context Protocol (MCP) integrated with FastAPI to prototype a Symptom Checker AI Assistant. The system connects a browser-based frontend with an MCP-powered backend, illustrating how agentic AI workflows can be modularized and extended with tools.

The project is designed as both a learning resource for MCP and a practical health AI prototype.

Architecture

Flow:
Browser → FastAPI → MCP Client ⇆ MCP Server (Tools) → Results

• Browser (Frontend): Collects user input.
• FastAPI (Backend): Handles HTTP routes, acts as entry point.
• MCP Client: Async bridge between FastAPI and MCP.
• MCP Server: Hosts modular tools (Extractor, Matcher, Advisor).
• Results Layer: Outputs structured health insights (Symptoms, Diseases, Advice).

The repository includes an animated architecture GIF visualizing the system flow.

Protocols and Standards Used

• HTTP/REST → Communication between Browser and FastAPI.
• Python async → For MCP client bridging.
• JSON-RPC over stdio → Protocol for MCP client-server communication.
• MCP (Model Context Protocol) → Standard for tool orchestration.

Technologies and Frameworks

• FastAPI (backend web framework, async-first).
• MCP (Model Context Protocol for AI tool orchestration).
• Python 3.10+ (core runtime).
• Matplotlib + NumPy (visualization and architecture animation).
• ImageIO (GIF generation).
• HTML5 video and images (frontend UI banner integration).

Features

• End-to-end MCP architecture demo with FastAPI.

• Browser-based input and results rendering.

• Modular MCP server tools (Extractor, Matcher, Advisor).

• Animated architecture diagram for clear visualization.

• Applied to a Symptom Checker prototype:

  • Symptom extraction
  • Disease matching
  • Advice generation
  • Web search with TAVILY

  ---

Repository Structure

Project Structure

This project is organized into the following key directories and files:

.
├── app.py                   # The main FastAPI application entry point.
├── mcp_client/              # Contains the implementation for the MCP client.
├── mcp_server/              # Houses the MCP server, including various tools and the LLM (Large Language Model) integration.
├── static/                  # Stores static frontend assets, such as images and video.
├── templates/               # Contains the HTML templates for the frontend.
├── mcp_architecture_gif.gif   # An animated visualization of the MCP architecture.
└── README.md                # The main README file for this repository.