What We're Building
The AI-Native Manufacturing Operations Management platform we're building on top of the production 19-module MFX operational core — one module at a time, on a Strangler Fig migration path that doesn't disturb what already runs.
This page reflects what we're building right now. It will evolve as the build does — that's the point.
What Runs Today
Eukodyne MFX is a 19-module Manufacturing Operations Management system spanning MES, TMS, LIMS, and Enterprise applications — grown one module at a time since 2008 and refined in production at process and hybrid manufacturers ever since. It covers planning, batch execution, quality, inventory, tank management, recipe management, shipping, loading, railcar, customer communications, and demand forecasting. The data model is the asset: hundreds of validated tables and thousands of stored procedures encoding decades of vertical-specific operational logic. That asset is what the AI-Native upgrade is built on top of.
What We're Building
Three layers, each landing incrementally on top of the production core.
MCP-Native API Tier
A .NET 10 minimal-API layer wraps the existing 19-module operational core and exposes every meaningful operation as an MCP tool with OpenAPI documentation, per-agent identity, audit logging, and idempotency. Agents and humans become co-equal callers.
Industrial Data Fabric
A UNS broker bundled with every install (Mosquitto + our GPU broker plugin), MQTT-native event streams across all modules, and a vector store over operational data — recipes, SOPs, exception archives, lot history — so agents can retrieve semantically, not just by key.
AI Surface
A planning copilot that reasons against tank state, schedule conflicts, and recipe constraints. A quality copilot that explains exception root causes against years of CoA history. SparkHub7 — an on-prem agentic chatbot reading the UNS in real time. Every surface backed by the MCP tier.
How We're Building It
Strangler Fig migration. The new .NET 10 API tier and React UI grow around the existing production system. Each migrated screen and each MCP-exposed operation lands incrementally; nothing in production gets ripped out. Customers running MFX today see the new surface arrive a module at a time — not as a forced cutover.
The 19-module operational core stays. The data model stays. The stored procedures stay. The surface modernizes, the API tier mediates, and the AI capabilities ride on top. This is the cheapest, safest, fastest path to an AI-Native platform for process and hybrid manufacturers who can't afford a 24-month rebuild gap.
Cloud-Deployable on Microsoft Azure
The new API tier and AI services are designed to deploy on Azure for cloud customers, while on-prem customers continue to run MFX on their own infrastructure. The platform supports both deployment models from the same codebase.
Azure OpenAI / AI Foundry
Planning copilot, quality copilot, agent orchestration
Azure App Service
.NET 10 API tier and React UI hosting for cloud customers
Azure SQL Managed Instance
SQL Server 2025 for cloud-deployed tenants with native vector search
Azure Container Apps
SparkBroker GPU MQTT broker and supporting services
Azure AI Search
Vector retrieval over operational data (RAG)
Microsoft Entra ID
Per-agent identity and authentication
Where We Are Right Now
An honest snapshot of build state as of May 20, 2026.
Want to Watch This Come Together?
Schedule a demo of what runs today, and we'll walk you through where each module is heading.