Agentic AI & MCP NOW
AI agents that don't just answer, but act: calling your APIs, querying your data and following your business rules. Built with LangChain4j and the Model Context Protocol, with guardrails, evaluation and full traceability.
We design and ship AI agents and cloud-native backends on the JVM — secure, observable and built to run in production, not just in a demo.
Our stack
What we do
Robust, scalable backend systems and intelligent applications — from cloud-native microservices to AI agents that act safely on your enterprise data and systems.
AI agents that don't just answer, but act: calling your APIs, querying your data and following your business rules. Built with LangChain4j and the Model Context Protocol, with guardrails, evaluation and full traceability.
Cloud-native Java with fast startup, low memory and native builds — ideal for containers and serverless.
Enterprise-grade backends on modern Java — virtual threads, records, pattern matching — designed to last.
Moving legacy Java EE and Spring applications to Quarkus and current Java versions — step by step, without a big-bang rewrite.
Containerised, observable services on AWS and OpenShift — with infrastructure as code, CI/CD and OpenTelemetry from day one.
How we work
We identify where AI genuinely adds value in your processes — and where it doesn't.
A working slice against your real data and systems in weeks, measured against clear success criteria.
Security, guardrails, evals, observability and cost control — everything a demo skips.
Running on your AWS or OpenShift platform, documented and maintainable by your own team.
// an AI agent with tools — Quarkus does the rest
@RegisterAiService(tools = OrderTools.class)
@McpToolBox("erp")
@InputGuardrails(PiiFilter.class)
public interface SupportAgent {
@SystemMessage("""
You are an enterprise support agent.
Use the tools. Never guess order data.
""")
String chat(@MemoryId String session,
@UserMessage String question);
}
Selected work
From regulated healthcare to our own product at sea — where security, standards and reliability are not optional.
Replace a legacy desktop application with a modern platform — running in the cloud and on appliances inside medical practices, without losing existing data.
Built end-to-end: Quarkus microservices on Java 25, web-component frontends and a central hub that syncs practices, forms and lab catalogues. Lab data exchange (LDT), standard forms with barcodes, a guided migration from the legacy system and AI-based suggestions of lab tests from diagnoses.
One codebase for cloud and on-prem, deployed via GitOps on OpenShift and monitored with Grafana dashboards and alerting.
Secure, standards-based messaging for healthcare organisations that meets the strict security requirements of a national healthcare network — operated in the cloud.
A high-security layer in Java and Quarkus in front of the Matrix homeserver: every request is authenticated and filtered, federation is limited to approved servers from a centrally maintained list, and participants are looked up in a FHIR directory. Plus self-service registration with Keycloak and OpenID Connect and a browser client built with web components.
A containerised messaging platform on an open standard, with full control over who can sign in, what may be called and which servers may federate.
Bring all of a luxury yacht's data together — instruments, AIS, energy, weather and cameras — on every device on board, and keep it working at sea without internet.
A Quarkus application on a compact onboard server reads NMEA 0183 and AIS, Victron energy data via MQTT and onboard sensors, and streams everything live to the browser on iPad, laptop and phone. Offline nautical charts, logbook, routes, anchor and watch alarms, weather and boat guard with push notifications.
Everything safety-relevant runs on the server and survives a restart — alarms sound in every browser and on the onboard speaker. A companion to certified navigation equipment, not a replacement.
Principles
What separates an impressive demo from a system your business can rely on.
Answers based on your documents and systems, with sources — not on model guesswork.
EU hosting, private models or on-prem via Ollama or vLLM — aligned with GDPR and the EU AI Act.
Evals, tracing and cost metrics make quality and spend visible — before and after go-live.
Native builds and lean JVM services start in milliseconds and keep cloud bills small.
Let's work together
Greenfield system, Quarkus migration or your first production AI agent — tell us about your project and we'll get back to you.