Product · How it works
From a raw tag to a value you can trust.
MaestroHub is the industrial context platform. It reads your machines and systems, gives every value a name, a unit, a quality check and a record of where it came from, and serves it to people, data platforms and AI tools. This page follows one machine through it, step by step.
01 What it is
One layer between your plant and everything that reads it.
And what it is not.
Every machine and system in a plant speaks its own protocol and names its data its own way. Without a layer in the middle, every dashboard, data platform and AI tool has to connect to each of them separately. MaestroHub is that layer: each system connects to it once, and everything else reads from one place.
Without a layer in the middle
With MaestroHub
Context Engine
Connects to your machines and systems, names and checks every value, and serves it over MQTT, REST and MCP. Includes approvals, access control and the audit trail.
Knowledge Graph
The plant as things and the links between them: what feeds what, and what stops if this stops.
App Studio
Apps a domain expert describes, built on the platform and run inside that person's permissions. With the Maestro agent inside the product.
- Not a control system.A data path, not a control loop. Your PLCs and SCADA keep running the machines.
- Not your long-term historian.It keeps 1 to 32 days per topic. Your historian, or the data platform you choose, keeps the rest.
- Not a replacement for MES or ERP.It reads from them and writes back to them, under permissions and audit.
02 The problem
A machine sends a number, not its meaning.
What OP20 actually sends, and what is missing.
This is what OP20, a CNC station, sends over OPC UA: an address inside the machine and a number. Nothing in it says what the number is, what unit it is in, whether it is fine, or which machine it came from.
ns=1;s=OP20-CNC.CoolantTemp64.8- What is it?A node ID, not a name anyone uses.
- Which unit?°C, °F, or a raw count.
- Is it fine?Nothing says what normal is.
- Where from?Which machine, which read, when.
Today every dashboard, report and AI tool that reads this value has to answer those four questions again, each in its own way. That is how one plant ends up with different numbers for the same thing. MaestroHub answers them once, where the value enters, and everyone reads the answer. The next section shows how, on this exact value.
03 From tag to meaning
How a raw tag gets its meaning.
One machine, OP20, a CNC station that speaks OPC UA. Five steps, in the product's own words.
Live3.0, Oct 2026
Nothing is guessed. Your engineer names each value once, where it enters MaestroHub, and everyone reads that name from then on.
Connect
Pick the connector for the machine, enter its address and test the connection. A connection that cannot reach the machine says why.
Connect › ConnectionsNew connectionOPC UAOP20-CNC · <host>:14840ConnectedBrowse
The machine describes itself. You open its address space and watch its values change as it works.
Connection › Address SpaceAddress SpaceObjects › OP20-CNCns=1;s=OP20-CNC.SpindleSpeedns=1;s=OP20-CNC.CoolantTempName it
Each value you pick becomes a function: a named, tested read you can reuse anywhere.
Connection › FunctionsFunctionsStateSpindleSpeedSpindleLoadCoolantTempPlace it
A pipeline reads the functions on a schedule, shapes the values and publishes them to one address in the Unified Namespace: company, site, area, line, machine. Every value in the plant gets one place, and everyone finds it there.
Orchestrate › PipelinesUNS topicacme companyberlin sitemachining arealine1 lineop20 machineA pipeline in the designer, on a demo plant. Each box is one step you can test on live data. Give it a contract
A schema says what the value is: its type, unit, valid range and alert thresholds. Bind it to the topic, and every publish is checked against it. Start from one of 88 templates for pumps, energy meters, OEE and more.
Explore › SchemasSchema · cnc.temperature- Type
- float64
- Unit
- °C
- Range
- 0 to 120
- Warn
- ≥ 70
The Schemas registry, with a schema bound to one topic.
The same value, before and after
Before
ns=1;s=OP20-CNC.CoolantTemp64.8
After
acme/berlin/machining/line1/op20coolant_temp · Coolant temperature64.8 °CGood
- What is it?Coolant temperature of OP20, at one address in the plant.Steps 3 and 4
- Which unit?°C, stated by the schema.Step 5
- Is it fine?Yes: inside 0 to 120, below the warning at 70.Step 5
- Where from?Read from OP20-CNC over OPC UA, with its quality, stored on the value itself.Automatic
Illustrative value on the simulator's OP20 station.
Every dashboard, data platform and AI tool now reads the value on the right. None of them has to answer the four questions again, so they all get the same number.
04 Three ways to do it
By hand, from a file, or with an assistant.
All three produce the same connections, pipelines and topics.
One machine at a time
The five steps above, in the product's screens, with a pipeline and a dashboard at the end.
A whole line at once
Import a .mhub.json file. You read a preview first, and pipelines arrive switched off until you turn them on. The counts are the simulator's file.
Describe it, then check it
Claude, Cursor or VS Code do the same steps through MaestroHub's MCP server, with your own access token. You can open and change every piece afterwards.
Start with one machine. You do not have to model the whole plant before it is useful. The first named value can be on a dashboard the same day; the rest of the line follows when you are ready.
05 Checking a value
One value, and how it was made.
When a number looks wrong, check it instead of asking around.
Because the source is stored on every value, anyone can open a value and walk back how it was made, right down to the machine it was read from.
- The machineOP20-CNCOPC UA connection
- The readCoolantTempfunction, read at 06:16:44
- The pipeline runOP20 Liverun of 06:16:44, owned by your team
- The value64.8 °C Goodline1/op20 · coolant_temp
Illustrative trail. Quality is set by the connector that read the value, and a computed value is never better than its worst input.
In the product, the Data Explorer draws the same walk for any value:
06 What it connects to
The machines and systems you already run.
More than 90 connectors reach over 120 kinds of machines and systems. 22 can also write back, under permissions and audit.
PLCs, CNCs and controllers
OPC UA, OPC DA, Siemens S7, S5 and LOGO!, Modbus, EtherNet/IP, Mitsubishi MELSEC, Omron, Beckhoff TwinCAT, FANUC, Keyence, Panasonic, Opto 22, MTConnect, BACnet, CANopen, IEC 104
Brokers and messaging
MQTT, Sparkplug B, Kafka, AMQP, RabbitMQ, NATS, LoRaWAN, TCP and UDP
Historians and SCADA
AVEVA PI System and PI Web API, Ignition, InfluxDB, Prometheus
ERP, MES and databases
SAP IDoc and SAP Event Mesh, SQL databases and ODBC, MongoDB, Redis, REST, SOAP, FTP and file shares
Cloud and data platforms
Snowflake, Databricks, BigQuery, Microsoft OneLake, Amazon S3, AWS IoT Core and SiteWise, Azure IoT Hub and Event Hubs, Google Pub/Sub
How you read it back out
A built-in MQTT broker on port 1883, a REST API, an MCP server for AI assistants, dashboards and alerts in the product, and any connector on the left as a destination.
07 Where it runs
Next to your systems, inside your network.
Your systems of record stay where they are.
MaestroHub runs on your own infrastructure. It reads from your existing systems, gives their data one meaning and passes it on. Nothing leaves your network unless you send it.
What you already run
- PLCs, CNCs, robotsOPC UA, S7, Modbus
- SCADA and historiansIgnition, PI System
- MES and ERPSAP, SQL, REST
- BrokersMQTT, Sparkplug B, Kafka
MaestroHub
Stores a working window: 1, 7, 14 or 32 days per topic, 1 day by default.
Who reads it
- Your peopleOperators, engineers, maintenance
- AI assistantsClaude or any MCP client
- Data platformsSnowflake, Databricks, Kafka
- Your historianPI System, InfluxDB as destinations
08 AI assistants
They can build. They cannot overstep.
Why named values matter for AI, and what keeps it in bounds.
Live3.0, Oct 2026
Because every value now has a name, a unit and an address, an AI assistant can work with your plant the way an engineer does: connect a machine, build a pipeline, chart a value. MaestroHub decides what it is allowed to do.
- 1
It acts as you
The assistant connects to the MCP server with a personal access token you create. It sees and does only what your role allows.
- 2
A person approves
An Approval step in a pipeline holds the data until someone approves it. Nothing is approved automatically; an unanswered request expires as rejected.
- 3
Every step is recorded
The audit trail is hash-chained, so an edited or deleted record breaks the chain, and one call checks it.
You, in ClaudeConnect OP20 over OPC UA, read spindle speed and coolant temperature every 5 seconds, publish them to acme/berlin/machining/line1/op20 and chart spindle speed.
connection created · OP20-CNCdone
4 functions created and testeddone
pipeline "OP20 Live" saved and enableddone
dashboard "My OP20" createddone
ClaudeDone. Spindle speed arrives every 5 seconds. Everything I made is in MaestroHub under your name.
Illustrative conversation, following the getting-started guide.
09 Your first day
See it on your own machine.
No factory needed: the free simulator plays one.
- Minute 1Download and start
One file for Windows, macOS or Linux. No licence key, nothing else to install.
- First minutesLive data
Start the Digital Factory Simulator, import its file, and a four-station line is streaming.
- About 30 minutesYour first machine
Connect one machine yourself, build its pipeline and a dashboard.
- Six weeksPilot on your plant
One virtual machine, your real data, no writes to production. Then you decide.
The free trial includes every feature, Enterprise ones included, in two-hour sessions you can restart as often as you like.
10 Questions
Asked plainly.
How does MaestroHub give a raw PLC tag its meaning?
An engineer names it once, where it enters. You connect to the machine, browse its address space, turn the values you need into named reads called functions, and a pipeline publishes them to one address in the Unified Namespace, such as acme/berlin/machining/line1/op20. A schema bound to that topic sets the type, unit, range and alert thresholds, and every publish is checked against it. Every value also carries its quality and where it came from.
Do I have to model the whole plant before MaestroHub is useful?
No. You can start with one machine. The guided first machine, with its pipeline and a dashboard, takes about 30 minutes, and the rest of the line can follow when you are ready. A whole line can also be brought in at once by importing a .mhub.json file.
Which protocols does MaestroHub support?
More than 90 connectors reach over 120 kinds of machines and systems, including OPC UA, Siemens S7, Modbus, EtherNet/IP, MQTT, Sparkplug B, Kafka, PI System, Ignition, SAP, SQL databases and REST, plus cloud and data platforms such as Snowflake, Databricks and Azure IoT Hub. 22 connectors can also write back, under permissions and audit.
Does MaestroHub replace my historian, SCADA or MES?
No. It is a data path, not a control loop, and it sits next to the systems you already run. It keeps a working window of history, 1, 7, 14 or 32 days per topic with 1 day by default, and long-term history stays in your historian or goes to the data platform you choose.
Can an AI assistant set MaestroHub up?
Yes. Claude, Cursor, VS Code and other MCP clients can connect a machine, create functions, build a pipeline and a dashboard through MaestroHub's MCP server. The assistant uses a personal access token you create, so it can never do more than your own role allows, and everything it builds can be opened and changed by hand.
Follow these steps on your own machine.
Download MaestroHub and the free Digital Factory Simulator, and name your first value today. Or start the six-week pilot on your own plant.