Fhir MCP Server (Momentum)
Healthcare interoperability standards enable LLM agents to perform FHIR resource operations.
Data last scanned today · Reviewed today
Overview
This server bridges AI agents and FHIR-compliant backends using the Model Context Protocol. It supports full CRUD operations on clinical resources, enabling natural-language interaction with medical data structures. The architecture includes RAG-ready pipelines for processing documents and semantic search capabilities via Pinecone vector storage. Built on the FastMCP framework, the implementation provides modular tools for FHIR resource management and LOINC terminology validation. Users can configure the server via environment variables and deploy using included Docker support.
Our verdict
With a high-risk security score of 29/100, this server requires caution as it provides full write access to FHIR resources without a read-only mode. As an community-maintained project that hasn't been editorially reviewed and lacks recent updates, it should be restricted to isolated development or testing environments rather than production healthcare systems.
- Supports full CRUD operations on FHIR resources
- Integrates RAG pipeline with Pinecone vector search
- Includes LOINC API for medical terminology resolution
- Zero direct dependencies reduces supply chain surface
- High security risk due to no read-only mode
- Not an official vendor-maintained project
- Inactive repository with no commits in over nine months
Tools
| Tool | Description | Risk |
|---|---|---|
| request_patient_resource | Manage patient demographic and administrative information | medium |
| request_observation_resource | Handle clinical measurements and assessments | medium |
| request_condition_resource | Manage patient problems and diagnoses | medium |
| request_medication_resource | Handle medication information and orders | medium |
| request_immunization_resource | Manage vaccination records | medium |
| request_encounter_resource | Handle patient visits and interactions | medium |
| request_allergy_intolerance_resource | Manage patient allergy information | medium |
| request_family_member_history_resource | Handle family health history | medium |
| request_generic_resource | Operate on any FHIR resource not covered by specific tools | medium |
| request_document_reference_resource | Manage FHIR DocumentReference resources | medium |
Compatibility
| Client | Local | Docker | Remote | Read-only |
|---|---|---|---|---|
| Claude Desktop | ||||
| Cursor | ||||
| VS Code | ||||
| Windsurf | ||||
| ChatGPT |
Frequently asked questions
›What are the core prerequisites to run the FHIR MCP server?
You need access to a FHIR API server like Medplum, a Pinecone API key for semantic search capabilities, and either Docker or uv installed for dependency management.
›What happens if I do not provide a Pinecone API key?
The server will function, but the semantic retrieval and AI-powered document search features will be unavailable.
›How is the server configured?
The server uses extensive configuration options managed through a .env file.
›Does the server support authentication with FHIR backends?
Yes, the server includes secure OAuth2 token management to facilitate integration with FHIR APIs.
›Can the server handle document ingestion?
Yes, it supports AI-powered ingestion and chunking for TXT, CSV, JSON, and PDF document formats.
Badge
Maintain this server? Add the live badge to your README.