The Industrial DataOps platform that connects, models, conditions and delivers OT data. Configured in a browser, deployed at the edge.
OPC UA input
ns=2;s=Pump01.Temp = 74.19Deadband 0.5 · quality good
74.21 suppressed (Δ 0.02)
76.90 forwarded (Δ 2.71)Model: Pump · validated
{ "asset": "pump-01",
"temperature": 76.9,
"unit": "°C" }MQTT v5 · Unified Namespace
acme/istanbul/packaging/line-3/pump-01
{ "temperature": 76.9, "unit": "°C" }Illustrative example
Most plants already collect the data. What is missing is a consistent way to give it meaning and move it where it is needed. HighByte Intelligence Hub sits between machines, historians, MES and databases on one side, and analytics, cloud and AI on the other. You define once what a pump, a line or a batch looks like, map every real source onto that definition, and publish the result as clean, contextualised data.
Every new dashboard, historian or AI project asks the same questions: which tag is the pump temperature, what unit is it in, who owns the mapping? When each project answers with its own scripts, the plant ends up with integrations nobody can safely change.
Industrial DataOps moves that work into one governed layer. Models describe the asset once, pipelines apply the conditioning rules once, and every consumer receives the same payload.
From raw source to ready-to-use data in five steps.
Add a connection for each source: OPC UA servers, MQTT brokers, SQL databases, REST APIs, files, the PI System. One connection can have several inputs and outputs, and SQL or REST requests can use values from other systems as parameters.
Describe an asset type once as a model: attributes, units, metadata, nested structure. Create an instance for each real machine and map every attribute to its source.
Pipelines read, filter, buffer, transform and validate. Apply deadband, aggregate fast values to a slower rate, alarm on bad or stale data, and validate every payload against its model before it leaves the hub.
Organise the result in a namespace; for a Unified Namespace use an ISA-95-style hierarchy. The embedded MQTT broker (v3.1.1 and v5, JSON and Sparkplug) can host it, or you can publish to any broker, historian or cloud target.
Analytics, MES, historians and cloud platforms subscribe or call the REST data server. AI agents reach the same governed data through the Industrial MCP Server, with pipelines exposed to them as tools.
The capability groups; the details are on the features page.
Connect devices, files, databases and systems through open standards and native connectors, configured in a code-free interface.
Represent machines, products, processes and systems with reusable templates, and merge data from several systems into one contextualised object.
Clean and shape values at the edge, before they reach anything else.
Move raw data, modeled data or files between connections on an interval, on an event, or when called.
Design a Unified Namespace or any destination with a hierarchical structure: MQTT topics, historian asset trees, SCADA/IIoT tag structures.
Expose the same governed data to applications over HTTP and to AI agents over MCP, and speed up configuration with AI.
The same software from a single-board computer to an industrial server, at the edge or in the cloud.
On gateways and industrial PCs next to the machine, so conditioning happens before data talks to the cloud.
A central hub in the plant data centre that collects the namespaces of remote hubs.
As a cloud VM or container, close to cloud targets such as Databricks, Snowflake, AWS and Azure.
Typical situations; all of them are on the use cases page.
PLCs, MES and the historian are linked point to point, and every new system means another integration.
Data scientists spend most of their time decoding tag lists and fixing units.
Batch information is scattered across MES, PLCs and quality systems and compiled by hand for reports.
We do not just resell the product: we deploy it and hand it over to your team.
In a workshop we clarify the current state, sources, goals and constraints, and choose the first use case together.
We set up a HighByte pilot with a limited scope and written success criteria, and judge the result against them.
We standardise the models, policies and templates from the pilot, integrate with surrounding systems and run a security review.
We train your team, hand over documentation and support you after go-live.
It connects through open standards and native connectors such as OPC UA, MQTT and Sparkplug, SQL, REST/HTTP, files and the PI System; version 4.5 added NATS and i3X client connections. The connector overview also lists Modbus TCP, Apache Kafka, CSV and Parquet files, InfluxDB, Aspen InfoPlus.21, Snowflake, AWS services and the Ignition module. See HighByte’s documentation for the exact connector list; we validate it against your sources together.
A historian stores time series, SCADA lets operators run the process, an ETL tool moves batch data. Intelligence Hub sits between them: it models and conditions the data and serves it to all of them in the same shape. It does not replace them, it feeds them.
Not necessarily. The hub includes an embedded broker with MQTT v3.1.1 and v5 and JSON and Sparkplug payloads, enabled from the UI. If you already run an enterprise broker, the hub can publish to that instead.
Yes. HighByte states that it runs at the edge on single-board computers, industrial switches, IoT gateways and industrial servers. It installs on bare metal, a virtual machine or a container, and 4.5 added arm64 images.
Through the Industrial MCP Server. The agent reaches modeled data in the hub and the pipelines you choose to expose as tools, not the raw sources. What each tool can access is configured and traceable in the hub.
Tell us where you are and what you want to achieve, and we will outline the right scope and pilot plan. Our team replies within 24 hours.
