I was inspired by the MCP project and decided to explore this space myself. At first, my goal was simply to understand what could be done with it and how it might be integrated with different AI tools. Later, I realized it would be valuable to rebuild it in PHP and extend it with functionalities that I, as a developer, felt were missing.
The result is my MCP Server, now available from: https://github.com/code24-nl/openehr-mcp-server.
Over the past week, I’ve tested it with Claude, LibreChat, and JetBrains AI, primarily using Sonnet 4. I must admit, it’s remarkable what these AI engines are now capable of. In addition to smaller refinements, I also added MCP Tools that allow querying archetypes from CKM as well as retrieving openEHR type specifications.
My motivation is closely related to a talk I’ll be giving at an upcoming meetup in the Netherlands. The topic: how developers can leverage MCP and AI to support their daily work with openEHR technology. I’ve been experimenting with use cases such as:
- letting agents/assistants generate code based on the Reference Model (RM) and/or archetypes,
- providing explanations of various openEHR concepts or models,
- assisting with design decisions (e.g., choosing the right archetype for a requirement),
- comparing, summarizing, or explaining RM aspects.
I strongly believe these capabilities could greatly support new developers as they begin working with openEHR, making the learning curve less steep and life easier.
Thanks, @odeak