Healthcare MCP Server

MCP (Model Context Protocol) server for healthcare data access via FHIR API. Built for the Prompt Opinion Agents Assemble Challenge.

Features

• Patient Summary - Get comprehensive patient overview (demographics, conditions, medications)
• Medications - List active medication requests with dosage instructions
• Lab Results - Retrieve recent laboratory observations
• Conditions - Browse patient conditions with clinical status

Standards

• FHIR R4 - HL7 Fast Healthcare Interoperability Resources
• SHARP-on-MCP - Standardized Healthcare Agent Remote Protocol
• MCP - Model Context Protocol by Anthropic

FHIR Context Support

Implements ai.promptopinion/fhir-context extension:

• X-FHIR-Server-URL header - FHIR server endpoint
• X-FHIR-Access-Token header - SMART on FHIR access token
• X-Patient-ID header - Current patient identifier

Required scopes:

• patient/Patient.rs (required)
• patient/Condition.rs
• patient/MedicationRequest.rs
• patient/Observation.rs

Installation

bun install

Build

# Build stdio server (default)
bun build src/index.ts --outdir dist --target bun

# Build HTTP server (Streamable HTTP)
bun build src/http-server.ts --outdir dist --target bun

Usage

Option 1: Stdio Server (Local)

bun dist/index.js

Test with:

echo '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}' | bun dist/index.js

Option 2: HTTP Server (Remote / Prompt Opinion)

TRANSPORT=http PORT=8080 bun dist/server.js

Server will be available at http://your-host:8080/mcp

Deployment

Deploy to Render (Recommended - Free Tier)

1. Push code to GitHub/GitLab
2. Create new Web Service on Render.com
3. Connect repository: yourname/healthcare-mcp
4. Settings:
   • Runtime: Docker
   • Environment Variables:
     • TRANSPORT = http
     • PORT = 8080
5. Deploy - server will be available at https://your-app.onrender.com/mcp

Or use the render.yaml file for Infrastructure as Code.

Deploy to Fly.io

# Install flyctl: https://fly.io/docs/hands-on/install-flyctl/
fly launch --dockerfile Dockerfile
fly deploy

Deploy to Railway

# Install Railway CLI: https://docs.railway.app/develop/cli
railway login
railway init
railway up

With Prompt Opinion

1. Deploy the HTTP server to a public URL (e.g., Render, Fly.io, Railway)
2. Add MCP server in Prompt Opinion: Configuration -> MCP Servers
3. Enter your server URL: https://your-server.com/mcp
4. Enable FHIR context extension when prompted
5. Authorize requested scopes

Test Your Deployment

# Test if server is running
curl -X POST https://your-server.com/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-11-25","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}'

Tools

| Tool | Description | |------|-------------| | get_patient_summary | Complete patient overview | | get_medications | Active medications list | | get_lab_results | Recent lab results | | get_conditions | Patient conditions |

Example FHIR Server

For testing, you can use:

• HAPI FHIR Server
• SMART Health IT Sandbox

Tech Stack

• TypeScript
• Bun runtime
• @modelcontextprotocol/sdk
• Zod validation

Project Structure

healthcare-mcp/
├── src/
│   ├── index.ts          # Stdio server (local)
│   ├── http-server.ts   # HTTP server (remote)
│   ├── fhir-context.ts  # FHIR context extraction
│   └── tools/
│       └── healthcare.ts # MCP tool implementations
├── dist/                # Build output
├── package.json
└── tsconfig.json

Testing with MCP Inspector

npx @modelcontextprotocol/inspector bun dist/index.js

License

MIT