In 2026 Ignition, CODESYS and OPC UA gained MCP servers, and Rockwell and Siemens showed engineering agents. Here is a tested MCP server for a production line in Python: reads from a tag allow-list, setpoint changes only after a person approves, and where such a server belongs in a plant.

OmniMES exposes an MCP (Model Context Protocol) endpoint — a bridge through which any AI assistant talks directly to your production system. We walk through, step by step, how to go from an empty system to a finished dashboard using a fictional WOODCRAFT furniture plant as an example — with no UI and not a single line of SQL.

Model Context Protocol (MCP), released by Anthropic in November 2024 and adopted in 2025 by OpenAI, Google and Microsoft, has become the de facto standard for connecting LLM agents to external data and tools. For the MES world, it ends the era of "one integration per model" and starts an architecture in which a plant exposes its data once — and Claude, ChatGPT, Gemini, and home-grown agents all plug in. This article explains how MCP works, what to expose from an MES, how to design it safely (lethal trifecta, IT/OT zoning), and when MCP simply does not make sense.