Patient Engagement MCP Server And Client

A sample healthcare-focused AI application demonstrating the Model Context Protocol (MCP) architecture.

The project includes:

• MCP Servers exposing healthcare-related tools
• MCP Client for communicating with MCP servers
• AI Agent (MCP Host)
• Flask-based web interface
• SQLite database for persistence

Architecture

User
  │
  ▼
Flask Web UI
  │
  ▼
AI Agent (MCP Host)
  │
  ▼
MCP Client
  │
  ├───────────────┐
  ▼               ▼
Patient MCP     Scheduling MCP
Server          Server
  │               │
  └───────┬───────┘
          ▼
      SQLite DB

Demo Screenshot

Project Structure

Patient-Engagement-MCP-Server-and-Client/
│
├── README.md
├── requirements.txt
├── .env.example
│
├── app/
│   ├── __init__.py
│   │
│   ├── web/
│   │   ├── __init__.py
│   │   ├── app.py
│   │   ├── templates/
│   │   └── static/
│   │
│   ├── agent/
│   │   ├── __init__.py
│   │   ├── agent.py
│   │   └── prompts.py
│   │
│   ├── client/
│   │   ├── __init__.py
│   │   └── mcp_client.py
│   │
│   ├── servers/
│   │   ├── __init__.py
│   │   ├── mcp_patient_data_server.py
│   │   └── mcp_scheduling_server.py
│   │
│   ├── db/
│   │   ├── __init__.py
│   │   ├── database.py
│   │   └── schema.sql
│   │
│   └── config.py
│
├── data/
│   └── patient_engagement.db
│
└── tests/
    ├── test_patient_server.py
    └── test_scheduling_server.py

Components

Web UI

Location:

app/web/app.py

Responsibilities:

• Accept user questions
• Display AI responses
• Communicate with AI Agent

AI Agent (MCP Host)

Location:

app/agent/agent.py

Responsibilities:

• Interpret user requests
• Select appropriate MCP tools
• Call MCP servers through MCP client
• Generate final responses

MCP Client

Location:

app/client/mcp_client.py

Responsibilities:

• Connect to MCP servers
• Discover tools
• Execute MCP tool calls
• Read MCP context resources

MCP Context Resources

The MCP servers expose reusable context through resources:

• context://patient-data/overview
• policy://registration/washington-only
• policy://registration/adult-only
• patient://{patient_id}/context
• context://scheduling/overview
• appointments://{patient_id}/context

The AI Agent loads the static context and policy resources during startup so tool selection and final responses can use server-provided context.

Patient Data MCP Server

Location:

app/servers/mcp_patient_data_server.py

Example tools:

• get_patient
• search_patients
• update_patient

Scheduling MCP Server

Location:

app/servers/mcp_scheduling_server.py

Example tools:

• get_appointments
• create_appointment
• cancel_appointment

Database Layer

Location:

app/db/database.py

Responsibilities:

• SQLite connection management
• CRUD operations
• Data persistence

Installation

Clone repository:

git clone https://github.com/nguyener/Patient-Engagement-MCP-Server-and-Client.git
cd Patient-Engagement-MCP-Server-and-Client

Create virtual environment:

python -m venv venv
source venv/bin/activate

Install dependencies:

pip install -r requirements.txt

Running the Application

Start Patient Data MCP Server

From project root:

python -m app.servers.mcp_patient_data_server

Start Scheduling MCP Server

From project root:

python -m app.servers.mcp_scheduling_server

Start Web Application

From project root:

python -m app.web.app

Open browser:

http://localhost:5000

Example Questions

Patient Queries:

• Show patient 1001
• Find patient John Smith
• Update patient email

Appointment Queries:

• Show appointments for patient 1001
• Schedule appointment for next Monday
• Cancel appointment 2005

Technologies

• Python
• Flask
• SQLite
• Model Context Protocol (MCP)
• OpenAI SDK
• HTML/CSS/JavaScript

Future Enhancements

• Authentication and authorization
• Notification MCP Server
• Email and SMS integration
• FHIR integration
• Oracle Health EHR integration
• Vector search and semantic patient lookup
• Multi-agent orchestration
• LangGraph workflow support