operational-data-mcp

A Model Context Protocol (MCP) server that provides real-time operational data visibility across industrial systems.

Built to unify fragmented telemetry from production lines, machines, and plant-level systems into a single, queryable interface — solving the problem where operational data exists but can't be accessed usefully across isolation boundaries.

The Problem It Solves

In large industrial facilities, every production line and machine often runs on isolated networks for reliability and security reasons. Management has no real-time view of what's actually happening:

• Downtime: Walk the floor with a clipboard to find out why a line stopped

• Throughput: No live production counts until end-of-shift reports

• Bottlenecks: Can't see which line is constraining overall plant output

• Data silos: Systems that were never designed to talk to each other

This MCP server unifies that telemetry.

Related MCP server: TrakSYS MCP Server

Use Cases

• Real-time production visibility: See live downtime, throughput, and efficiency metrics across lines

• Anomaly detection: Identify unusual production patterns automatically

• Operational dashboards: Feed Claude with live plant data to generate insights

• Compliance & audit: Generate operational evidence for certifications (STIG, cATO)

• Predictive maintenance: Use historical patterns to flag equipment issues early

How It Works

The server exposes five core tools:

1. list_lines — List all production lines and their data availability

2. get_plant_summary — Plant-wide uptime, production totals, line count

3. get_line_status — Current status of a specific line (running/down, recent output)

4. summarize_throughput — Efficiency metrics for a line (uptime %, total output, average output/hour)

5. find_anomalies — Detect production anomalies using statistical deviation detection

Installation

npm install operational-data-mcp

Running the Server

npx operational-data-mcp

The server listens on stdin and writes JSON-RPC responses to stdout. It's designed to be invoked by Claude via MCP.

Integrating with Claude

Add this to your Claude configuration:

{
  "mcpServers": {
    "operational-data": {
      "command": "npx",
      "args": ["operational-data-mcp"]
    }
  }
}

Then Claude can access operational data:

User: "What's the plant-wide uptime right now?"
Claude: [calls get_plant_summary] → Returns average uptime across all lines

User: "Find any anomalies on Line 1"
Claude: [calls find_anomalies with lineId: "line_1"] → Returns statistical outliers

Sample Data

The server ships with realistic sample data: a 72-hour dataset from a 2-line ice cream carton packing facility, with:

• Hourly production output per line

• Simulated downtime events

• Statistical anomalies to detect

Use this to test the tools without needing a real plant connection.

Integration Pattern

In a real deployment, you would:

1. Connect to your actual industrial data sources (PLC networks, API gateways, data historians)

2. Normalize the data into the same schema the MCP server expects

3. Serve it via the same tool interface

4. Claude gains instant access to live operational intelligence

Why This Matters

Most enterprises have operational data trapped in silos:

• Multiple vendor systems that don't interoperate

• Network isolation for security that blocks visibility

• No unified query interface

• Complex, bespoke integrations for every new use case

This MCP server is a pattern for how to surface that data in a way that:

• Respects security and network boundaries

• Unifies fragmented sources into a single interface

• Works with AI agents that can ask sophisticated questions

• Scales across new data sources without code changes

Author

Lee Marcum

Built to prove that industrial-grade operational visibility is achievable with clean abstractions and the right protocol.

License

MIT