2026 ASP Dijital AI Architecture Series

Industrial AI with MCP Architecture

Integrating AI into Production Systems via Model Context Protocol

Prepared by: ASP Dijital Dönüşüm Hizmetleri A.Ş. • Date: June 2026 • Version: 1.0

Executive Summary

Model Context Protocol (MCP) is an open protocol that enables AI agents to communicate with external systems in a standardized way. In industrial environments, MCP bridges OPC UA servers, databases, and APIs with LLM-based agents.

This report explains step-by-step how to design industrial AI systems using MCP, how to pull real-time data from OPC UA, and how to use LLM agents for production optimization.

Why MCP? Barriers to Industrial AI

Barrier	Traditional Approach	MCP Solution
Data Silos	Manual ETL pipelines	Live MCP server connection
Protocol Diversity	Custom code per protocol	Standardized MCP interface
Security	Hardcoded credentials	Centralized auth & token management
Scalability	Monolithic integration	Modular MCP server architecture
Maintenance	Fragile custom connectors	Standardized, documented API

MCP Architecture Layers

🤖 LLM Agent (Claude, GPT-4, Local Model)

↓ MCP Client

🔌 MCP Host (Application / Orchestrator)

↓ MCP Protocol (JSON-RPC)

📡 MCP Server (OPC UA Connector)

↓ OPC UA / Modbus / MQTT

🏭 PLC / SCADA / Sensors

OPC UA MCP Server Setup

The MCP server exposes the OPC UA address space for natural language queries:

// MCP Server: opcua-bridge
// Tools: read_node, write_node, browse_namespace, subscribe_data

import { Server } from '@modelcontextprotocol/sdk/server/index.js';
import { OPCUAClient } from 'node-opcua-client';

const server = new Server({
  name: 'industrial-opcua-bridge',
  version: '1.0.0',
});

server.tool(
  'read_node',
  'Read OPC UA node value',
  async (params: { nodeId: string, endpoint: string }) => {
    const client = new OPCUAClient();
    const session = await client.connect(params.endpoint);
    const data = await session.readVariableValue(params.nodeId);
    return { value: data.value, timestamp: data.sourceTimestamp };
  }
);
            

Use Cases

1. Natural Language Production Queries

User: "What is the average temperature of Furnace 3 over the last 2 hours?"

MCP Agent: Finds OPC UA node → fetches data → computes → responds

2. Anomaly Detection & Alerting

Agent continuously monitors pressure sensor values. On threshold breach:

Sends notification to relevant operator
Automatically queries valve position
Triggers safety protocol

3. Maintenance Assistant

Technician: "When was Pump 7 last serviced? And what is its current vibration level?"

Agent: Combines maintenance record from CMMS + real-time sensor data from OPC UA.

Expected Performance Gains

45%

Query Resolution Time Reduction

24/7

AI-Assisted Operations

30%

Operator Workload Reduction

<500ms

Avg Response Time

Security & Compliance

MCP servers use TLS 1.3 encrypted connections
Every agent query is logged to audit trail
OPC UA authentication (X.509 certificates) is preserved
LLM prompt injection protection is applied
Data minimization: only necessary nodes are exposed
Rate limiting controls system load

Implementation Roadmap

Week 1-2: Identify critical OPC UA servers, set up MCP server POC
Week 3-4: LLM agent integration, basic query scenarios
Week 5-6: Security layer, audit logging, rate limiting
Week 7-8: Pilot deployment on production line
Week 9-12: Scale-out, additional scenarios, training

Next Steps

Explore the MCP server demo environment
Request an OPC UA address space MCP readiness assessment
Consult ASP Dijital AI team for LLM selection
Define pilot project scope