We're excited to announce the development of a new Model Context Protocol (MCP) server specifically designed for FHIR tooling! 🚀

What is MCP?

The Model Context Protocol is Anthropic's open standard that enables AI assistants to securely connect to external data sources and tools. It creates a bridge between AI models and the rich ecosystem of development tools, databases, and APIs that developers use daily.

Our FHIR Tooling Vision

Healthcare interoperability requires robust tooling, and we're building an MCP server that will provide AI assistants with native access to FHIR operations. Our goal is to make working with FHIR resources as natural as working with any other data format.

Starting with FHIRPath

For our first implementation, we're focusing on FHIRPath - the powerful expression language for navigating and extracting data from FHIR resources.

We'll be leveraging our comprehensive FHIRPath implementation in Rust that features:

🔧 Comprehensive function support: 99+ FHIRPath functions across 10 categories

📊 Full specification compliance: Collection operations, string manipulation, date/time functions, math operations, type checking, and more

🏗️ Modular architecture: 11 specialized crates for maximum flexibility

🔄 Advanced features: Environment variables, Bundle resolution, custom functions, and CDA extensions

💾 Memory efficient: Arc-based resource sharing and arena memory management

Function Categories We Support

Our implementation includes comprehensive support for:

• Collection Operations: where(), select(), first(), last(), take(), skip(), union(), distinct(), and more

• String Functions: contains(), startsWith(), endsWith(), matches(), replace(), split(), upper(), lower()

• Date/Time Operations: now(), today(), timeOfDay(), component extraction functions

• Math Functions: abs(), ceiling(), floor(), round(), sqrt(), ln(), exp(), power()

• Type Operations: is(), as(), ofType(), type conversion functions

• FHIR-Specific: resolve(), extension(), conformsTo(), children(), descendants()

• Conversion Functions: Type checking and conversion across all FHIRPath types

• Utility Functions: trace(), defineVariable(), encode(), decode(), and more

What This Means for Developers

With our MCP server, AI assistants will be able to:

Patient.name.where(use = 'official').family

Bundle.entry.resource.where(resourceType = 'Observation').value

Patient.birthDate < %today - 18 'years'

All with the reliability of our battle-tested Rust implementation that powers production healthcare systems.

Repository and Development

The MCP server development is happening at: Github Repo

We're building this as an open-source project, and we welcome contributions from the FHIR community. Whether you're working on healthcare applications, clinical research, or health IT infrastructure, this tooling will help streamline your FHIR workflows.

What's Next

Stay tuned for updates as we progress through:

1. Core FHIRPath MCP server implementation

2. Additional FHIR validation tools

3. Resource transformation utilities

4. Clinical terminology services integration

Follow our progress on GitHub and join us in building better tooling for healthcare interoperability!

---

This project builds on our existing FHIRPath implementation which has been refined through extensive testing against the official HL7 FHIRPath specification.