# the-momentum/fhir-mcp-server [Health: Active]

**Category:** 🧬 Biology & Bioinformatics  
**Repository:** https://github.com/the-momentum/fhir-mcp-server  
**GitHub Stars:** 101  
**Views:** 4  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/the-momentum-fhir-mcp-server

## Description
MCP Server that connects AI agents to FHIR servers. One example use case is querying patient history in natural language.

## Tools
Capabilities this server exposes over MCP:

- **Security** — OAuth2 authentication and proper access control for all operations

## Claude Desktop Quick Installation
No confirmed install details yet, so no config is published. Follow the setup steps in the repository above.

## Documentation

## What the-momentum/fhir-mcp-server MCP server does

The the-momentum/fhir-mcp-server MCP server gives MCP-compatible AI clients a structured interface to FHIR-compliant healthcare systems. It is intended for workflows where an agent needs to retrieve, create, update, or delete FHIR resources while communicating through natural-language requests. The repository demonstrates use with Medplum and Claude Desktop, including patient history questions and treatment-related queries.

The project covers major FHIR resource types with dedicated tools for selected resources and a generic tool for others. The documented examples include patient, condition, medication, and encounter requests. It also connects to LOINC services for medical terminology lookup and validation.

## How it works

The server uses FastMCP to expose MCP tools to a client. A FHIR client component manages communication with the configured FHIR API and obtains OAuth2 credentials for protected operations. When an agent asks about clinical information, the client can select resource-specific tools or the generic resource tool based on the request.

Document workflows use ingestion and chunking for TXT, CSV, JSON, and PDF files. Processed content can be embedded and searched through Pinecone, supporting semantic retrieval and retrieval-augmented generation workflows. Pinecone is required for document search; without it, those retrieval features are unavailable. An optional LOINC account enables access to current codes from the official API, while static or inferred codes remain available without that account.

## Setup and configuration

The the-momentum/fhir-mcp-server MCP server can run with Docker or uv. Setup requires cloning the repository, copying `config/.env.example` to `config/.env`, and supplying the required credentials and configuration. Docker execution uses `make build`; uv-based execution uses `make uv`.

The server can use stdio, HTTP, or HTTPS transport. A FHIR server account is required. Pinecone credentials are needed for semantic document search, and LOINC credentials are optional.

## Tools and capabilities

- Perform CRUD operations on FHIR resources.
- Request patient, condition, medication, and encounter resources with dedicated tools.
- Use a generic resource request tool for additional FHIR resource types.
- Search LOINC codes through the terminology service integration.
- Ingest and chunk TXT, CSV, JSON, and PDF documents.
- Run vector-based semantic document searches through Pinecone.
- Authenticate FHIR API operations with OAuth2.

## Limitations and notes

The the-momentum/fhir-mcp-server MCP server requires access to a FHIR API and does not replace the permissions enforced by that backend. Document search depends on Pinecone, and LOINC results may rely on static or inferred codes when an official LOINC account is not configured. The README presents synthetic FHIR data generation as an experimental use case rather than the project’s primary design target. The repository is categorized under Biology, Medicine and Bioinformatics, so deployments should be evaluated against the operator’s clinical-data governance and access-control requirements.

_Full upstream README: https://allmcps.com/mcp/the-momentum-fhir-mcp-server/readme_