mcpField
H
Healsens FHIR MCP

Healsens FHIR MCP

Local
healsens.comHealth1Updated 28w ago

About Healsens FHIR MCP

Healsens FHIR MCP is a read-only server for teams connecting AI agents to FHIR R4 or R5 systems. It combines embedded conformance definitions with the live server CapabilityStatement to guide typed reads, searches, and supported operations.

Key features

  1. 01FHIR R4 and R5 support
  2. 02Typed resource reads and searches
  3. 03Live capability discovery with caching
  4. 04Embedded conformance definitions
  5. 05Read-only operation invocation
  6. 06stdio and SSE hosting

Use cases

  1. 01Retrieve resources from a configured FHIR endpoint
  2. 02Search using FHIR-aware parameters
  3. 03Inspect endpoint capabilities and schemas

MCP support

Exposes FHIR read, search, discovery, and read-only operation tools, plus schema and guide resources to MCP clients. Runs over stdio or SSE.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

No install method listed yet

We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.

Repository

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

healsens-fhirmcp

Open-source, conformance-aware MCP server for FHIR R4 and R5.

healsens-fhirmcp is Healsens' MCP implementation for healthcare systems that already expose FHIR and want to make that data usable by AI agents safely, predictably, and with far better schema awareness than a thin proxy can provide. It fronts a configured FHIR base URL and exposes a focused, read-only MCP surface for resource read, typed search, conformance discovery, and safe read-only operations.

The core idea is simple: models perform better when the server helps them understand the target FHIR API. This project embeds version-matched FHIR conformance artifacts and combines them with the live CapabilityStatement of the connected server so tool calls stay grounded in the shape, semantics, and capabilities of the actual endpoint.

Why This Project Exists

Most healthcare AI integrations fail in predictable ways: generic HTTP wrappers leak too much surface area, models guess at search parameters, and FHIR differences between servers turn otherwise simple tasks into brittle prompt engineering. healsens-fhirmcp is designed to solve that problem directly.

It gives teams a practical MCP layer that is:

  • FHIR-aware instead of endpoint-agnostic
  • grounded in real conformance metadata instead of prompt-only assumptions
  • safe by design with a read-only toolset and a fixed target server
  • usable in production integration flows over stdio or native SSE

This repository is open sourced to give implementers a concrete, credible foundation for healthcare MCP deployments, not just a toy example.

Demo Access

A live demo is available for teams evaluating the project. To request access, email contact[at]healsens[dot]com.

What It Provides

  • Single configured FHIR target server
  • Read-only MCP tool surface
  • FHIR R4 and R5 support
  • Embedded core StructureDefinition and OperationDefinition registries
  • Live CapabilityStatement fetch with caching
  • Typed read and typed search
  • Conservative read-only operation invocation
  • Discovery and schema resources under the fhirmcp:// URI scheme
  • Native stdio and SSE hosting modes

MCP Tools

  • fhir_read
  • fhir_search
  • fhir_capability_statement
  • fhir_structure_definition
  • fhir_operation_definition
  • fhir_invoke_operation

MCP Resources

  • fhirmcp://server/guide
  • fhirmcp://server/overview
  • fhirmcp://server/capability-statement
  • fhirmcp://schema/structure-definitions
  • fhirmcp://schema/structure-definition/{selector}
  • fhirmcp://schema/operation-definitions
  • fhirmcp://schema/operation-definition/{selector}
  • fhirmcp://server/search-guide

Design Principles

  • Conformance first: the server uses embedded FHIR artifacts together with the live target CapabilityStatement to shape tool behavior.
  • Read-only by design: this implementation is intentionally focused on safe retrieval, discovery, and supported read-only operations.
  • Predictable integration surface: one configured target, explicit schemas, bounded behavior.
  • Real-world interoperability: supports both FHIR R4 and R5, including server-specific capability discovery at startup.

The project deliberately does not try to be a full FHIR SDK, a generic proxy, or a write-capable orchestration layer.

Quick Start

  1. Copy examples/config.yaml and set target.base_url to your FHIR R4 or R5 server.
  2. Export the bearer token environment variable named in the config if your server requires authentication.
  3. Sync the vendored conformance artifacts.
  4. Build and run the server over stdio or SSE.
./scripts/sync_fhir_core_defs.sh
go mod tidy
go build -o bin/healsens-fhirmcp ./cmd/healsens-fhirmcp
FHIRMCP_CONFIG=./examples/config.yaml ./bin/healsens-fhirmcp

With server.transport: stdio, the binary behaves as a traditional subprocess MCP server over stdin and stdout. With server.transport: sse, it exposes a native SSE MCP endpoint at http://127.0.0.1:8081/sse by default.

SSE Mode

Set the transport block in examples/config.yaml:

server:
	transport: sse
	listen_address: 127.0.0.1:8081
	sse_path: /sse
	cors:
		enabled: true
		allowed_origins:
			- http://localhost:8788

Then start the server:

go build -o bin/healsens-fhirmcp ./cmd/healsens-fhirmcp
FHIRMCP_CONFIG=./examples/config.yaml ./bin/healsens-fhirmcp

Your MCP SSE endpoint will be:

http://127.0.0.1:8081/sse

If the client runs in a browser context, server.cors controls cross-origin access for the SSE endpoint and its POST message endpoint. allowed_origins should contain the exact browser origins you trust.

Development

  • make sync-core refreshes the vendored FHIR R4 and R5 artifacts.
  • make build builds the server binary for stdio or SSE mode.
  • make test runs unit and integration tests.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.