Medplum MCP Server

Demo MCP server that proxies requests to Medplum, allowing you to use the Model Context Protocol (MCP) with Medplum's FHIR data.

What is Model Context Protocol (MCP)?

MCP is an open protocol that standardizes how applications provide context to LLMs. Think of MCP like a USB-C port for AI applications. Just as USB-C provides a standardized way to connect your devices to various peripherals and accessories, MCP provides a standardized way to connect AI models to different data sources and tools.

Setup

Create a .env file with MEDPLUM_BASE_URL:

# To use Medplum hosted:
MEDPLUM_BASE_URL=https://api.medplum.com/

# To use a local Medplum server:
MEDPLUM_BASE_URL=http://localhost:8103

By default, this MCP proxy server runs on port 8104. You can change the port by setting the PORT environment variable in your .env file:

# Change the port to 5000
PORT=5000

Run dev server

To run the development server, use the following command:

npm run dev

Run production server

First, build the production server:

npm run build

Then, start the production server:

nohup node --max-old-space-size=8192 dist/main.cjs &

Testing with MCP Inspector

Start the inspector:

npx @modelcontextprotocol/inspector

Testing Streamable HTTP

Set "Transport Type" to "Streamable HTTP" (recommended transport).

Set "URL" to the /stream path on your server, e.g. http://localhost:8104/stream.

Testing SSE

Set "Transport Type" to "SSE" (required by Claude and ChatGPT).

Set "URL" to the /sse path on your server, e.g. http://localhost:8104/sse.