Book a consultation

Technology cluster

HiveMQ: MQTT Broker, Platform, Cloud and Edge Explained

HiveMQ explained for plant engineers: the broker, Cloud, Edge and 2026 Platform suite on the ISA-95 stack, with a release and lifecycle model.

Vladimir Romanov


HiveMQ is an MQTT broker and the family of industrial data software built around it. Plants meet it as the message hub between machines, gateways, SCADA, historians and cloud systems, most often as the broker at the centre of a Unified Namespace. This page is a map for engineers: what the HiveMQ portfolio contains, where each piece sits on the ISA-95 stack, how its releases and support windows work, and where to read more. It is current as of 21 November 2026 and built from HiveMQ’s own releases, documentation and pricing page, plus Eclipse and GitHub sources. Where HiveMQ describes itself, we say “HiveMQ states”. Where a view is ours, we say so.

This is an independent technical profile. HiveMQ is not a Joltek partner. For the company itself, its history, funding, leadership and competitors, see HiveMQ GmbH, the company behind it.

HiveMQ in brief

HiveMQ states that its company, then called dc-square, was founded in 2012 in Germany, and that the HiveMQ broker began as a small part of a consulting business (HiveMQ 4 release, 12 December 2018). The company overview holds the full history. The short version for an engineer is a sequence of technology moves.

MQTT 5 in the broker, 12 December 2018. HiveMQ 4 shipped, and HiveMQ states it fully implements MQTT 5. Whether it was the first broker to do so we have not checked, so we do not claim it.

Open source and a new name, April and May 2019. HiveMQ launched Community Edition and its Java MQTT client as open source on 16 April 2019, and renamed the company from dc-square to HiveMQ on 27 May 2019 (rename post).

Edge, Data Hub and managed Cloud, 2023 to 2024. HiveMQ Edge, an open-source gateway that turns industrial protocols into MQTT, was announced in July 2023, and Data Hub, which validates and transforms data inside the broker, in September 2023. HiveMQ Edge reached general availability with commercial features on 19 March 2024.

From broker to platform, 2025 and 2026. HiveMQ announced Pulse, a data intelligence layer over MQTT, as a private preview on 18 February 2025. On 9 September 2026 it announced a suite also named HiveMQ Platform, covering Connect, Contextualize, Analyze and Act. Whether the suite’s Contextualize and Analyze layers are the former Pulse is not confirmed in anything we read, so we do not say they are.

HiveMQ portfolio on the ISA-95 stack

The portfolio is easiest to understand by where each piece runs. The broker and its data features live at level 3 and in the cloud. The Edge gateway lives at the machine side, at levels 1 and 2, translating what PLCs speak into MQTT. HiveMQ makes no PLCs, SCADA or historians, so the diagram marks those as not HiveMQ products. Placement is our reading of HiveMQ’s documentation, not a HiveMQ reference architecture.

The HiveMQ portfolio placed on the ISA-95 stack Five horizontal bands from the cloud and level 4 at the top to level 0 process and sensing at the bottom. At the top, HiveMQ Cloud and the HiveMQ Platform suite layers Contextualize, Analyze and Act. At level 3, the Enterprise MQTT Broker as a cluster with Data Hub and extensions to Kafka, databases and cloud services. At level 2, the open-source Community Edition broker, the MQTT client library and the MQTT command line tool, with SCADA and historians that are not HiveMQ products as MQTT clients. At level 1, the HiveMQ Edge gateway with protocol adapters, a local broker and an MQTT bridge. At level 0, PLCs, drives and sensors, which are not HiveMQ products and speak OPC UA, Modbus, Siemens S7 or EtherNet/IP. Arrows show data flowing up through the Edge bridge to the central broker and on to the cloud. Where each HiveMQ product runs, from the machine to the cloud Cloud and level 4 Enterprise, analytics, AI Level 3 · Operations Site or central message backbone Level 2 · Supervisory SCADA, HMI, historian, area brokers Level 1 · Control edge Gateway next to the controllers Level 0 · Process and sensing The machine, the tag, the signal HiveMQ Cloud · Platform suite layers Managed brokers. Contextualize, Analyze, Act (public preview). Enterprise MQTT Broker · Data Hub · extensions Clustered broker, in-broker validation, Kafka and database links. Community Edition · MQTT Client · MQTT CLI Open-source broker and tools. SCADA and historians are not HiveMQ. HiveMQ Edge Protocol adapters, local broker, MQTT bridge upstream. Not HiveMQ products: PLCs, drives, sensors OPC UA · Modbus TCP · Siemens S7 · EtherNet/IP · BACnet/IP A typical placement, not a HiveMQ reference architecture. Data moves up as MQTT through bridges. SCADA and historians are third-party clients of the broker.
HiveMQ sells the transport and the data handling in the middle of the stack. The controllers below it and the SCADA and historians beside it come from other vendors, and the broker is the contract between them.

Read the diagram from the bottom up. An Edge gateway polls or subscribes to a PLC and publishes MQTT. A broker, at the plant or centrally, carries it. Data Hub and the extensions shape and forward it. Everything above that is about reaching and governing the data, not about control. In our view this separation is the design’s strength: a plant that loses its cloud link keeps its local broker and its level 2 clients running. See our MQTT in manufacturing guide for the general case.

The portfolio as a table

ProductWhat it isRuns onLicence
HiveMQ Enterprise MQTT BrokerThe clustered, commercial broker, sometimes called HiveMQ Platform in older documentsYour servers, VMs or KubernetesCommercial, tiered plans
HiveMQ Community EditionThe open-source technical core of the brokerA single JVMApache 2.0
HiveMQ CloudThe managed broker serviceHiveMQ’s hostingSubscription
HiveMQ EdgeA gateway that converts industrial protocols to MQTTLinux, Windows, macOS, containersApache 2.0, plus commercial features
Data Hub and extensionsIn-broker validation and transformation, and links to Kafka, databases and cloud servicesInside the Enterprise brokerIncluded in self-managed plans per the pricing page
HiveMQ Platform (suite)Connect, Contextualize, Analyze and Act (the documentation adds an assistant, Bea)HiveMQ’s consolePer HiveMQ’s pricing page
Glowing blue network of connected nodes and lines on a dark background
Many connections meeting at one broker: the design point of MQTT, and the reason the broker becomes shared infrastructure.

The HiveMQ release and lifecycle model

HiveMQ runs two release rhythms side by side, and a plant has to choose which to follow. Monthly minor releases arrive about the first week of each month and carry new features. Long Term Support (LTS) releases come far less often and receive fixes without feature churn. HiveMQ’s support guidance describes an LTS release about every 18 months, with monthly releases in between (HiveMQ support blog). Each month HiveMQ also ships a maintenance build of the LTS line, so a plant on the LTS still receives fixes.

The 2026 change that matters most: HiveMQ 4.55, released 1 September 2026, replaced 4.28 as the LTS release. HiveMQ’s 4.28.29 changelog, as we read it, continues 4.28 support for six months after the new LTS, through February 2027 (4.28.29). The same 4.55 release notes state that the system requirements remain on Java 21 as the minimum runtime and that Java 25 is supported and recommended (4.55 release notes). It also removed the first version of the Control Center web interface, so a plant that still uses Control Center v1 must move to v2.

HiveMQ release and support timeline, June 2026 to February 2027 A timeline from June 2026 to February 2027 with five rows. Monthly broker releases 4.53 on 30 June, 4.54 on 4 August, 4.55 on 1 September, 4.56 on 29 September and 4.57 on 3 November. The 4.55 long term support line runs from 1 September 2026 onward. The 4.28 previous long term support line is supported through February 2027. Java 21 remains the minimum runtime and Java 25 is recommended in 4.55. HiveMQ Cloud Serverless is available until it retires on 31 December 2026. A vertical line marks 21 November 2026. Jun Jul Aug Sep Oct Nov Dec 26 Jan 27 Feb 27 Monthly minors 4.53 4.54 4.55 4.56 4.57 30 Jun 4 Aug 1 Sep 29 Sep 3 Nov Current LTS 4.55 LTS from 1 Sep 2026, the line to track for few upgrades Previous LTS 4.28 LTS, maintenance builds continue; support through Feb 2027 Java runtime 4.55: Java 21 stays the minimum, Java 25 recommended Cloud Serverless HiveMQ Cloud Serverless available Retires 31 Dec 2026 21 Nov 2026: this page Dates from HiveMQ's changelog and Cloud documentation. The 4.28 end date and the Serverless date each rest on one HiveMQ page.
Two dates in the next three months decide most plans: Cloud Serverless ends on 31 December 2026, and 4.28 support ends in February 2027. A plant still on 4.28 has a 4.55 upgrade to schedule.

The lifecycle as a table

ItemWhat HiveMQ statesDateWhat it means for a plant
Current monthly releaseBroker 4.573 November 2026New features arrive monthly; test before production
Current LTS4.55 replaced 4.281 September 2026The line for plants that upgrade rarely
Previous LTS4.28.x support continues six months after the new LTSThrough February 2027Plan the move to 4.55 before then
Java runtimeJava 21 minimum (unchanged), Java 25 supported and recommended4.55Check JVM images, patching and the base OS
Control CenterVersion 1 removed, version 2 carries Data Hub viewsFrom 4.55Retrain operators on the new interface
Cloud ServerlessRetires, clients blocked, data deleted31 December 2026Move to another Cloud plan or a self-managed broker
EdgeJDK 25 from release 2026.82026.8Container and Helm users need no action; check ZIP installs
Community EditionYear-numbered releasesThe GitHub page pointed to 2026.6 on 21 November 2026 (no release date shown)Check the project page for the current one

In our view, the monthly cadence suits a Cloud tenant or a pilot, and the LTS line suits a plant broker. A broker is shared infrastructure for many systems, and every restart disconnects clients, so a plant is better served by fewer, planned upgrades than by monthly ones. Whether HiveMQ’s LTS cadence of about 18 months holds is something to watch rather than rely on. We could not confirm the exact 4.28.0 release date, so we do not give it.

The name “HiveMQ Platform” now means two things. Older documents use it for the broker, and on 9 September 2026 HiveMQ used it for a suite of Connect, Contextualize, Analyze and Act (HiveMQ announcement). The platform documentation also names an assistant, Bea, that the launch release does not mention. Act, which HiveMQ describes as agentic AI for operations, is a public preview, not generally available. When you read a quote or a specification, ask which one is meant. HiveMQ’s own documents use “Enterprise MQTT Broker” for the broker.

Choosing a HiveMQ product by job

Most plants do not need the whole portfolio. They need one or two pieces, chosen by the job, and the names make that harder than it should be. The table below maps a situation to the part of the portfolio that addresses it, using only what HiveMQ’s documentation says each part does. It is a starting point for the deeper pages, not a recommendation.

JobPart of the portfolioWhy it fitsWhat to check
Learn MQTT or run a bench testCommunity Edition, the MQTT CLI, or the free Lab planOpen source or free, and quick to startLab is limited to 25 connections, per the pricing page
Bring PLC tags onto MQTTHiveMQ EdgeAdapters for OPC UA, Modbus TCP, Siemens S7, EtherNet/IP, BACnet/IP and othersWriting back to a tag needs a commercial southbound mapping
A site broker that must stay upEnterprise MQTT Broker as a clusterMasterless cluster with replicated persistent dataDefault replica count is 2; plan node count and failure limits
One namespace across several plantsPlant brokers bridged to a central brokerBridging is HiveMQ’s documented multi-site patternBridge extension is commercial; test loop prevention and topic mapping
Stream plant data into Kafka or a databaseExtensions on the Enterprise brokerBidirectional Kafka and several database extensionsConfirm which extension and version your plan includes
Reject bad payloads at the brokerData HubSchemas, policies and transformations inside the brokerFree mode allows one policy; confirm limits for your plan
Avoid running serversHiveMQ CloudHiveMQ operates the brokerServerless retires 31 December 2026; Starter and Enterprise remain on the pricing page
Sparkplug devices and a primary applicationA Sparkplug-aware broker setupHiveMQ lists Sparkplug support and an open Sparkplug Aware extensionAsk which edition and version was tested against the Sparkplug specification

Three habits that prevent most wrong purchases

Ask for the edition and the version, every time. The same product name can mean a free build, a commercial plan or a hosted service, and features move between them. HiveMQ’s own pricing page lists self-managed plans by connection and throughput caps, and its Cloud documentation still lists older Serverless names. A quote that names only “HiveMQ” is incomplete. Our editions and pricing page shows how to read the tiers.

Separate the broker from what you do with the data. A broker moves messages. Contextualising them against an ISA-95 model, governing them and running AI on them are separate layers, in HiveMQ’s suite and in anyone else’s. A plant can run MQTT well for years with only the first. In our view the common failure is to buy the upper layers before the topic structure and the edge connections are sound, and then to blame the software for what is a design problem. Our guide to the Unified Namespace covers the design side.

Size for the failure case, not the average. The number of connections and the message rate in the plan are caps, and a plant should test what happens near them: a site reconnecting after a network outage sends a burst of queued messages and session restores at once. HiveMQ’s documentation describes an overload protection mechanism that slows a client down by stopping reads from its connection when the broker is under pressure. That protects the broker, and it also means a slow or overloaded broker shows up in your gateways as delayed data, not as an error. Our edge solutions guide discusses buffering at the edge for that reason.

How to read HiveMQ’s own material

HiveMQ publishes a lot: documentation, a changelog, a blog with detailed articles on MQTT, Sparkplug and Unified Namespace design, case studies and press releases. They carry different weight, and it helps to know which is which.

  • The changelog and documentation are the source for versions, dates, limits and defaults. We rely on them for every date on this page.
  • The pricing page is the source for plans and prices, and it changes. Note the date you read it.
  • Press releases and the platform page describe what HiveMQ intends and how it positions itself. We treat them as HiveMQ’s statements.
  • Case studies show what customers do and often carry few numbers. One names a large multi-factory deployment with a stated message volume. We read that as HiveMQ’s account, not an audit, and we say so wherever we use it.
  • The blog is educational and useful, though its articles are written by HiveMQ staff and tend to favour the HiveMQ approach, as vendor blogs do.

None of this is a criticism. It is the same discipline we apply to any vendor, and it is why this cluster separates “HiveMQ states” from “in our view”.

System index

HiveMQ Platform

The Enterprise MQTT Broker, the 2026 suite, clustering, deployment and the editions that run it.

HiveMQ Cloud

The managed service: Starter, Enterprise, the Serverless retirement and the choice against self-hosting.

HiveMQ Edge

The open-source gateway: protocol adapters for PLCs, bridging and what needs a licence key.

Extensions and Data Hub

Kafka, security and database extensions, and in-broker validation of data.

Unified Namespace

Using HiveMQ as the broker of a plant namespace, with Sparkplug and the Pulse history.

Open source

Community Edition, the MQTT client and the command line tools, and what each leaves out.

Alternatives

EMQX, AWS IoT Core and the other brokers, compared on the same dimensions.

Manufacturing

Ignition, plant network security and multi-site architecture.

Most asked

The three questions that bring most readers to a HiveMQ page, answered in depth:

  • Editions and pricing: Lab, Launch, Run and Enterprise as listed on 21 November 2026, what each includes, and the Cloud meter.
  • HiveMQ vs Mosquitto: when a lightweight open-source broker is enough and when a plant needs a clustered one.
  • Sparkplug: what HiveMQ’s Sparkplug support does and does not cover, and the Eclipse compatibility listing.

What HiveMQ is not

Boundaries help as much as features. HiveMQ’s products move and govern MQTT data, and a plant design still needs the parts HiveMQ does not make.

It is not a PLC, a SCADA or a historian. HiveMQ Edge reads from controllers and a broker carries what is published, but nothing in the portfolio runs control logic, draws an operator screen or stores years of process history as its main job. A SCADA and historian layer remains, and it connects to the broker as a client. Our data historian guide explains what that layer does that a broker does not.

It is not a replacement for the controls network. MQTT carries information between systems. It is not a fieldbus and should not carry safety or hard real-time control. The plant network architecture still decides what can connect to what, and the gateway sits on it as one more device to patch, monitor and segment.

It is not the whole Unified Namespace. A broker is one component. The namespace is also the naming, the payload structure, the ownership of each topic and the rules for who may publish. HiveMQ publishes guidance on each, and its Data Hub can enforce schemas, but the design decisions stay with the plant. That is the reason our own work centres on architecture rather than on any one product.

It is not a single product with a single price. The portfolio spans free open-source tools, self-managed commercial plans and a managed service, and each has its own terms. The system pages that follow take them one at a time so that a reader can find the terms that apply.

What HiveMQ means for a plant

These are our views, drawn from the public record above. They are meant for the person deciding whether and how to put a HiveMQ broker at the centre of a plant data design.

Choose the broker for ten years, not for the pilot. A free Community Edition or a Lab plan is a fine place to learn MQTT. A plant namespace outlives the pilot, so the questions that matter are clustering, upgrade policy, support terms and how data leaves the broker. Clustering and high availability and the editions page cover those in turn.

The broker is shared infrastructure, so treat it like a PLC. Every system that publishes or subscribes depends on it. That argues for change control, a tested restore, a documented topic structure and a segmented network. Our guides to OT network segmentation and ISA/IEC 62443 explain why. HiveMQ states certifications for its own organisation, such as ISO/IEC 27001 and SOC 2 Type II; those describe the vendor’s practices, not your plant’s deployment.

A gateway is not a historian. HiveMQ Edge can poll a PLC and publish changes, but the choice of polling interval, change-only publishing and store-and-forward behaviour decides whether the data is complete enough for a record. The PLC scan cycle and polling guide explains the underlying timing, and the HiveMQ Edge page lists which features sit behind a licence key.

Treat suite and AI claims as claims. The 2026 Platform suite is new, and Act is a public preview. HiveMQ’s statements about what the layers do are vendor statements, so test them on your own data, and keep a plan that works if a preview feature changes. A broker with a clean namespace and good governance is useful without any of them.

Compare on the same dimensions. HiveMQ is one of several capable brokers. Our event broker comparison and the HiveMQ alternatives page set out where each fits. We do not declare a winner, because the answer depends on scale, existing skills, hosting policy and budget.

Where Joltek works

We design MQTT and Unified Namespace architectures and integrate brokers with PLCs, SCADA, historians and enterprise systems as part of IT/OT projects. We are independent of HiveMQ: we have no commercial relationship with the company and do not resell its products.

That scope is deliberate. We do not claim to be a HiveMQ partner, a certified integrator or a source of HiveMQ support. If your question is about licences, support contracts or pricing, the answer comes from HiveMQ. If it is about how a broker fits a plant architecture, a network or a commissioning plan, that is our work, and it sits under IT/OT integration and the owner’s engineer role. For the wider context, see our pages on event-driven architecture and plant data, or contact us.

Limits and open questions

A fair profile names what we could not establish, as of 21 November 2026:

  • The Serverless retirement date has one source. It comes from HiveMQ’s Cloud documentation, and we found no second page that states it.
  • The 4.28 support end rests on one changelog entry. We read “through February 2027” from the 4.28.29 changelog as summarised, and did not find a standalone support policy page for the exact date.
  • Pulse and the suite are not linked in public documents. HiveMQ’s Operator 2.3.0 notes show Helm values named “HiveMQ Data Intelligence” replacing “HiveMQ Pulse”, but we found no sentence stating which suite layer is the former Pulse.
  • Sparkplug certification detail is thin. The Eclipse list shows HiveMQ Community Edition, Professional Edition and the Sparkplug Aware extension, without a version or date column in the page we read. Ask HiveMQ which edition and version was tested.
  • Per-adapter licensing on Edge is unclear. The documentation says writing to any tag needs a commercial southbound mapping, but the open or commercial status of each adapter is not stated in one place.
  • Prices change. The figures in this cluster are as listed on 21 November 2026 and should be re-read on HiveMQ’s pricing page before you rely on them.

Trademark notice. All product names, logos and brands are the property of their respective owners. Use of these names does not imply any affiliation with, endorsement by, or partnership with the manufacturer. Joltek is an independent consultancy; commercial relationships, where they exist, are disclosed on our Partners page.

Frequently asked questions

What is HiveMQ?

HiveMQ is an MQTT broker and the industrial data software built around it, made by HiveMQ GmbH. The portfolio has a self-managed Enterprise MQTT Broker, an open-source Community Edition, the HiveMQ Cloud service, the HiveMQ Edge gateway, Data Hub and extensions, and, since 9 September 2026, a suite also called HiveMQ Platform. Plants use it as the message backbone between machines, SCADA, historians and enterprise systems.

Which HiveMQ broker version is current, and which is the long term support release?

As of 21 November 2026, HiveMQ's changelog lists broker 4.57 (3 November 2026) as the current monthly release. HiveMQ 4.55 (1 September 2026) replaced 4.28 as the Long Term Support release. HiveMQ's documentation says 4.28.x support continues through February 2027. A plant that wants few upgrades should track the LTS line.

What Java version does HiveMQ need?

HiveMQ's 4.55 changelog states that the system requirements remain on Java 21 as the minimum version and that Java 25 is supported and recommended. HiveMQ Edge runs on JDK 25 from release 2026.8. Container and Helm users get the runtime inside the image and usually need no action.

Is HiveMQ Cloud Serverless being retired?

HiveMQ's Cloud documentation states that HiveMQ Cloud Serverless retires on 31 December 2026. After that date clients cannot connect, and clusters and data are deleted. HiveMQ's documentation says affected users can claim a free migration licence for the self-managed Enterprise Broker. We found no second source for the date, so confirm it with HiveMQ.

Is HiveMQ free?

Parts of it are. HiveMQ Community Edition and HiveMQ Edge are Apache 2.0 open source, with some Edge features behind a licence key. HiveMQ's pricing page, as listed on 21 November 2026, shows a free Lab plan limited to 25 connections, Launch at $299 per month and Run at $499 per month, with Enterprise on request.

Is Joltek a HiveMQ partner or reseller?

No. Joltek is an independent consultancy with no commercial relationship with HiveMQ and does not resell its products. We design MQTT and Unified Namespace architectures and integrate brokers with PLCs, SCADA, historians and enterprise systems, and we write about HiveMQ from public sources.

References and further reading

  1. HiveMQ changelog · HiveMQRelease dates for the broker, Edge and Platform Operator.
  2. What's new in HiveMQ 4.55 · HiveMQThe LTS release and the Java 21 minimum.
  3. HiveMQ 4.28.29 released · HiveMQThe 4.28 support window.
  4. Managing HiveMQ releases and support: what you need to know · HiveMQThe LTS and monthly release model.
  5. HiveMQ launches new Platform for manufacturers · HiveMQThe 9 September 2026 announcement.
  6. HiveMQ Cloud documentation · HiveMQThe Serverless retirement notice.
  7. HiveMQ pricing · HiveMQPlans and prices as listed on 21 November 2026.
  8. HiveMQ Edge documentation · HiveMQ
  9. HiveMQ Community Edition · GitHubApache 2.0 source and README.
  10. Sparkplug compatible software · Eclipse FoundationThe HiveMQ entries on the Sparkplug list.
  11. HiveMQ 4 released · HiveMQThe 12 December 2018 release with MQTT 5 support.
  12. A new name for dc-square · HiveMQThe 27 May 2019 rename.

← All technologies

Have a project that needs technical ownership?

Send us the situation. If it is not something we should take on, we will say so.

The Plant-Side Brief

The weekly brief for the people who own a plant’s control systems.

One email every Wednesday for engineering and operations leaders responsible for controls, SCADA, MES and OT networks. Skim the top in two minutes; read the Deep Dive when you have fifteen.

  • One modernization problem, worked through to cost, risk and a schedule you can defend. The Deep Dive.
  • End-of-support dates before they reach your plant Windows, servers, controllers and HMI software. Lifecycle Watch.
  • A checklist your engineers can run that week not a framework to admire. Field Kit.
  • Every number sourced, no vendor agenda we are paid by plants, never by vendors.