olivierb123/hybrid-ai-mcp-localhealthcoach

Local Health Context MCP Server

This project implements a local MCP (Model Context Protocol) server that exposes a tool (get_patient_background) to an Azure AI Foundry hosted agent.
It enables a cloud-based medical “specialist” agent to call into your private local LLM to fetch personal health context without sending sensitive data to the cloud.

This demonstrates a true Hybrid AI pattern:

• Sensitive data stays entirely on your machine
• The cloud agent handles reasoning and diagnosis
• A local LLM (via Foundry Local) provides private background
• A Dev Tunnel exposes your MCP server securely
• Azure AI Foundry orchestrates tool calling

Architecture Overview

User -> Cloud Agent (Azure AI Foundry) -> (MCP Tool Call) -> Dev Tunnel -> Local MCP Server -> Local GPU LLM (Foundry Local)

Workflow:

1. User reports symptoms to the cloud agent
2. The agent calls the MCP tool to request personal background
3. The MCP server loads your local patient_profile.json
4. It sends a prompt to your local GPU LLM
5. The anonymized summary is returned to the cloud agent

Repository Structure

src/
  mcp-local-health.py # MCP server + GPU inference logic
  patient_profile.json # User’s private medical history
  requirements.txt            # Python dependencies
README.md

Setup

1. Create and activate a virtual environment

Windows: python -m venv .venv ..venv\Scripts\activate

macOS/Linux: python3 -m venv .venv source .venv/bin/activate

2. Install dependencies

pip install -r requirements.txt

Configure Your Local Medical Profile

Edit the file patient_profile.json.

Example:

{
  "chronic_conditions": ["mild asthma"],
  "medications": ["albuterol inhaler"],
  "recent_labs": {
    "A1C": 5.4,
    "Vitamin D": 32
  }
}

This file remains local and is never transmitted.

Run the MCP Server

python mcp-local-health.py

You should see: [MCP] Listening at http://0.0.0.0:8081

Expose the MCP Server Using Dev Tunnels

Create the tunnel

devtunnel create mcp-health

Add port 8081

devtunnel port create mcp-health -p 8081 --protocol http

Host the tunnel

devtunnel host mcp-health

You will get a public URL such as: https://abcd1234.usw3.devtunnels.ms:8081

Connect the Tool in Azure AI Foundry

Open your agent → Tools → Add Tool → MCP.

Fill in:

Save — the cloud agent now calls your local LLM tool.

License

MIT License