MakeIntelligent Request a pilot

Machine data from the PLC to your AI, without leaving the plant.

MakeIntelligent reads your PLCs, stores the history on a server inside your plant, and lets Claude, Cursor or any MCP client ask questions about it. Your roles decide what the AI can see.

The plant hub

Every machine in. Every answer out.

Site Runtimes collect from the floor. The hub names, stores and governs it. Your AI and your data platforms take it from there, without a single PLC socket leaving the plant.

01Your floor57 connectors

  • S7
  • Logix
  • MELSEC
  • FINS
  • Modbus
  • OPC UA
  • FOCAS
  • ADS
See all drivers →

mii plant hub

  1. ModelISA-95 tree
  2. Historyplant-local
  3. Rulesalarms, events
  4. MCPgoverned access

02Ask over MCP

  • Claude, Cursor or any MCP client
  • Your own agents, read-only by default

03Publish through Uplinks

  • MQTT
  • Azure IoT Hub
  • Kafka
  • Sparkplug B

One product, three jobs.

You don't need a driver package, a historian and an AI gateway from three vendors. MakeIntelligent does all three, and the history never leaves your plant unless you send it.

How a pilot runs

  1. Connect the machines

    Native drivers for Siemens S7, Modbus TCP/RTU, EtherNet/IP, OPC UA, Omron and Mitsubishi. Set up devices, scan groups and tags in the Console, then test the connection before you commission.

  2. Keep the history in the plant

    Every sample lands in a plant-local historian. Operators get trends, ISA-18.2 alarms and dashboards. If you need the data elsewhere, Uplinks publish it to MQTT, Sparkplug B, Kafka, Azure IoT Hub or AWS IoT.

  3. Let AI ask the plant

    A built-in MCP server turns tags, assets, history and alarms into tools. Add one URL to Claude or Cursor and ask "why did Line 2 slow down?" The model never gets a connection to a PLC.

What gets installed at each level.

MakeIntelligent follows the Purdue model your OT team already uses. Nothing is installed on PLCs, and nothing below the DMZ talks to the internet. Pick a level to see what runs there.

The plant, as tools for your AI.

The plant hub includes an MCP server. Claude, Cursor or your own agent see 14 typed tools, and every answer traces back to a tag, a time range and a quality flag.

  • Find and read tags by asset or name, with unit and quality
  • Pull history from the plant historian without exporting CSVs
  • Walk the asset graph to answer "what stops if this pump stops?"
  • Check health of devices, runtimes and active alarms
  Claude, connected to mii 14 tools
Why did Line 2 slow down in the last hour?
mii_see_find("line 2")
12 tags on Packer-2, Conveyor-2B, Filler-2
mii_history_series("Conveyor-2B.motor_amps", 1h)
peak 18.4 A at 14:12, baseline 11.2 A, quality good
mii_understand_impact("Conveyor-2B")
feeds Packer-2, which feeds Palletizer-2
Conveyor-2B's motor current jumped 64% at 14:12, which points to a jam. It feeds Packer-2, so that explains the drop in throughput.

Speaks the protocols on your floor.

PLCs, CNCs, robots, meters and building systems — the same drivers you pick in Console. We name the exact model and firmware in the pilot contract.

Siemens S7
Allen-Bradley
Mitsubishi MELSEC
Omron
Modbus TCP/RTU
OPC UA
EtherNet/IP
Beckhoff ADS
FANUC FOCAS
CODESYS V3
ABB
IEC 61850 · 104

All 57 drivers and Uplinks, and which run on Lite

Start with one line and one question.

We connect one line, put its history on Trends, and give your team an MCP URL. The pilot succeeds when someone gets a useful answer from the plant.