openehr-assistant

The OpenEHR Assistant MCP Server provides a bridge between MCP clients and openEHR domain resources. It exposes a set of tools for CKM (Clinical Knowledge Manager) lookups, terminology resolution, and openEHR guides and type specifications, along with prompts to orchestrate multi-step workflows. Client applications can invoke tools such as ckm_archetype_search, ckm_archetype_get, ckm_template_search, terminology_resolve, guide_search, guide_get, and type_specification_search/get to perform archetype exploration, template discovery, terminology mapping, and guideline access. Prompts like ckm_archetype_explorer or guide_explorer help guide the AI through complex tasks by combining multiple tool calls, enabling tasks such as semantic explanations, language translations, and syntax corrections within openEHR contexts. The server is accessible remotely via a public endpoint (for production deployments) or locally via the streamable HTTP and stdio transports for development and testing.

Prerequisites:

• Docker installed on your machine (Docker Desktop or equivalent)
• Optional: the MCP client you plan to use for testing (e.g., Claude Desktop, Cursor, LibreChat)

Installation steps:

1. Pull the MCP server image from the container registry (GitHub Packages or Docker registry used by the project): docker pull cadasto/openehr-assistant-mcp

2. Run the MCP server (example using a port exposed for local access): docker run -p 8080:8080 -i cadasto/openehr-assistant-mcp

3. Verify the server is up by checking logs or hitting the MCP endpoint with your client:

   • Confirm the server is listening on the configured port (default 8080 in this setup).
   • Use an MCP client to enumerate available MCP elements or invoke a test tool, such as ckm_archetype_search or terminology_resolve.

4. (Optional) Configure environment variables for production:

   • PORT: port to expose (default 8080)
   • TRANSPORTS: transports to enable (e.g., streamable_http,stdio)
   • LOG_LEVEL: log verbosity (e.g., info, debug)

Notes:

• The project is in pre-release, so you may encounter breaking changes as features evolve.
• If you build locally, you can also use a local Dockerfile or a CI-produced image if provided by the project.

Tips and considerations:

• This MCP server targets openEHR workflows (archetypes, templates, terminology, guides, and type specifications). Use CKM-related tools to explore and fetch definitions, then feed those results into MCP clients for task automation.
• The server supports streamable HTTP and stdio transports for development; production deployments typically use a standard HTTP transport over a stable endpoint.
• If you customize prompts, ensure they reference available tools accurately (e.g., ckm_archetype_search with appropriate query parameters).
• Monitor logs for updates during pre-release phases, as endpoints and tool signatures may evolve before v1.0.
• For Docker deployments, consider running behind a reverse proxy and enabling TLS in production.