Book a consultation

HiveMQ

HiveMQ Extensions and Data Hub: The Catalogue and How It Fails

HiveMQ extensions explained: the SDK model, all 29 catalogue entries by licence class, Data Hub in brief, how to select them and how upgrades break them.

Vladimir Romanov


HiveMQ extensions are Java modules that run inside the HiveMQ broker and add what the MQTT core leaves out: who may connect, where messages are copied to, what gets measured. The extension model is why a broker that does one job can also write to Kafka, check a JSON Web Token and publish metrics. It is also where upgrades and licences go wrong, because an extension is code that you now have to keep in step with a frequent release cycle. This page covers how the model works, the catalogue by licence class, Data Hub in brief, how to select, and how it fails. It is current as of November 2026 and built from HiveMQ’s documentation and catalogue.

It sits under the HiveMQ hub and next to the HiveMQ Platform page. Three deeper pages are listed at the end. For the company behind the product, see HiveMQ the company.

How HiveMQ extensions work

An extension is a Java module loaded by the broker. It runs in the same process as the broker and talks to it through a published SDK. HiveMQ’s documentation says the open framework supports interception and modification of MQTT messages, integration with other services, statistics collection and fine-grained security. The development page requires OpenJDK 21 or higher and says Java 25 is also supported.

There are two SDKs. The Community Extension SDK is open and carries the core features above. The Enterprise Extension SDK adds cluster communication and reliable, scalable processing of incoming messages, and includes a signature module that HiveMQ says protects the integrity of your code. HiveMQ describes the Enterprise SDK as adding features for its professional and enterprise editions.

HiveMQ’s Kinesis extension documentation says all HiveMQ Enterprise Extensions are preinstalled in the release bundle and disabled by default. You enable the ones you have rights to. HiveMQ’s extension documentation says pre-built community extensions are available on GitHub and the HiveMQ website.

Hand placing a wooden block onto a stepped stack of blocks
Each extension is another block on the same stack: easy to add, and every block has to stay level when the base moves.
Where a HiveMQ extension sits between MQTT clients and outside systems MQTT clients connect on the left to the broker core in the middle. Inside the broker process, three extensions plug into the core through the extension SDK: a security extension, a Kafka extension and a custom extension. Arrows leave the extensions toward an identity service, a Kafka cluster and a database on the right. MQTT clients PLCs, gateways, apps HiveMQ broker process (Java) Broker core sessions, subscriptions, queues, cluster Extension SDK (Community and Enterprise) Security enterprise Kafka enterprise Your own your code Identity service Kafka cluster Any system Illustrative layout
Extensions run inside the broker process and reach outside systems from there. That is why an extension upgrade is a broker upgrade, and why a faulty extension is the broker's problem.

The HiveMQ extensions catalogue

HiveMQ’s catalogue lists 29 extensions as read on 20 November 2026. The table groups them by the class the catalogue gives. Counts are ours, from the list.

GroupExtensionsClass as listed
Streaming and cloud messagingKafka (bidirectional), Google Pub/Sub, Amazon Kinesis (bidirectional)Enterprise, commercial
DatabasesMicrosoft SQL Server, MySQL, PostgreSQL, MariaDB, CockroachDB, TimescaleDB, MongoDBEnterprise, commercial
Data platformsSnowflake, Data LakeEnterprise, commercial
OperationsDistributed TracingEnterprise, commercial
Broker topologyBridge (broker to broker)Enterprise, commercial
Access controlSecurity (roles, users, permissions; SQL, LDAP, JWT, OAuth, OIDC realms)Enterprise, commercial
ObservabilitySplunkCommercial, but shown in the community section: the catalogue contradicts itself
Metrics and monitoringPrometheus, InfluxDB, CloudWatch, HeartbeatCommunity, Apache 2.0
Access hardeningFile RBAC, Deny Root WildcardCommunity, Apache 2.0
Diagnostics and SparkplugMQTT Message Log, Sparkplug Aware, Sparkplug InfluxDBCommunity, Apache 2.0

That is 16 enterprise or commercial entries and 13 community entries. Two points deserve care. First, the Splunk entry carries both labels in HiveMQ’s own catalogue, so confirm its class with HiveMQ before you budget. Second, we found no Azure Event Hubs extension in the catalogue, so we do not say one exists.

On cost, HiveMQ’s pricing page, as listed on 20 November 2026, says all enterprise extensions are included in every self-managed plan: Lab, Launch, Run and Enterprise. The limits sit in the plan, such as connection counts, not in a per-extension fee. The editions and pricing page covers the plans.

Data Hub in brief

Data Hub validates and shapes data inside the broker. Data policies match a topic filter, check payloads against a JSON Schema or Protobuf schema, and apply an action on failure, such as dropping the message with a log entry or rerouting it. Behaviour policies watch client behaviour, with three predefined models in HiveMQ’s documentation: MQTT Events, Publish Duplicate and Publish Quota. Transformations use JavaScript. It is a broker feature, not an entry in the extensions catalogue, which is why it has its own page: HiveMQ Data Hub.

Licence note: HiveMQ’s documentation says free mode, included since 4.17, allows one policy with a limited feature set, and a 5-hour trial removes the limits. HiveMQ’s pricing page lists in-stream data governance in every self-managed plan. On HiveMQ Edge, Data Hub is a commercial feature behind a licence key.

Selecting an extension

Work through this in order.

  1. Name the job. Authenticate clients, copy data to a system, watch the broker, or reshape messages. Reshaping messages is Data Hub’s job before it is an extension’s.
  2. Check the class. Commercial, Apache 2.0 or the contradictory Splunk case. Write the answer in the design file with the date.
  3. Prefer the catalogue to custom code. A catalogue extension is maintained by HiveMQ against each release. Your own is not.
  4. Check direction and scale. Kafka and Kinesis are listed as bidirectional. Whether a given database extension reads back is for its own documentation to say.
  5. Decide where the data should leave. If the real need is analytics, the Kafka extension page covers when a stream is the right target. If the real need is who may publish where, see the Enterprise Security Extension.

The wider design question, which broker feeds which system, is a Unified Namespace question. Our Unified Namespace broker guide covers it without tying it to one vendor.

How HiveMQ extensions fail

These are field failure modes, and the reasoning is our view unless a source is named. The table is the short version.

FailureTypical triggerWhat to check before it happens
Version driftMinor release, LTS change, Java version changeRebuild and test every extension in the upgrade rehearsal
No code ownerContractor leaves after go-liveNamed owner, repository, build and test pipeline
Licence surpriseExtension enabled that the plan does not cover, or a mislabelled classDated entitlement list in the design file
Broker couplingSlow extension code in the broker processLoad test with the extension active
Silent data pathA database or cloud extension copies data off siteReview against the segmentation plan

Extension version drift on upgrade. HiveMQ has shipped minor releases about monthly lately (4.55 on 1 September 2026, 4.57 on 3 November 2026). Version 4.55 of 1 September 2026 became the current LTS. HiveMQ’s extension documentation says that since 4.53 several Enterprise SDK services and methods are deprecated and will be removed in the next LTS release. In our view, a plant that rebuilt its custom extension once and never again is the plant that finds this out during an upgrade window. HiveMQ’s 4.55 release notes say the Enterprise Extension SDK compile target rose from Java 11 to Java 21, so custom extensions built against the latest SDK need Java 21 or later to compile.

Ownership of custom extension code. An extension you write is an application. It needs a repository, a build, tests against the broker version, and an owner who still works there in three years. In our view, the usual failure is a contractor-written extension with no owner after go-live. If the capability exists in the catalogue, use it. If it does not, write down who owns the code before you write the code.

Licence surprises. Three come up. A commercial extension ships preinstalled in the bundle, so what you can switch on may be more than what you are entitled to run. Confirm entitlement in writing. The class label of an extension can look inconsistent within the catalogue, Splunk being the visible case. And Data Hub’s free mode is one policy, so a design with ten data policies is licensed work, not a default. We read this as a reason to put the entitlement list in the architecture document, dated.

Broker coupling. An extension that blocks or slows inside the broker is a broker problem, because it runs in the same process. In our view, keep extension logic short, push heavy work to the system behind it, and load-test with the extension on, not in the lab without it.

What we do here

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. Within that, when we assess a broker design that uses extensions we would:

  • list every extension and its class, with the date and source of each licence statement
  • ask for the upgrade rehearsal that includes the extensions, not only the broker
  • ask who owns any custom extension code and where it is built and tested
  • check what leaves the plant through a Kafka, database or cloud extension against the OT network segmentation plan
  • compare the choice with a plain bridge or Kafka Connect style option before accepting broker-resident logic

If that is a question your project has, see IT/OT integration.

Limits and open questions

  • We read the catalogue on 20 November 2026 and did not install or test any extension. Class labels are HiveMQ’s.
  • The edition requirement for Data Hub on the broker, beyond free mode and the trial, is not stated on the Data Hub page we read.
  • Whether Azure Event Hubs works through the Kafka extension is untested by us.
  • Which Enterprise SDK services are removed in the next LTS is in HiveMQ’s 4.53 upgrade guide, which we did not read. Read it before you plan.
  • The catalogue changes. Re-check it at purchase, and again at every major upgrade, because a class label or an SDK deprecation can change between releases.
  • We did not find a HiveMQ statement on how long a given extension build stays compatible across minor releases. Ask HiveMQ for its compatibility policy, in writing.

Deeper

For the platform these plug into, see HiveMQ Platform.

Frequently asked questions

What are HiveMQ extensions?

HiveMQ extensions are Java modules that run inside the broker process and add behaviour the core does not have: authentication and authorisation, bridges to other systems, message interception and statistics. HiveMQ publishes a catalogue of 29 of them, some commercial and some open source under Apache 2.0. You can also write your own against the extension SDK.

Are HiveMQ extensions free?

Some are. HiveMQ's catalogue lists a group as community extensions under Apache 2.0, for example Prometheus, InfluxDB and Sparkplug Aware. The Kafka, database, Security and Bridge extensions are enterprise extensions. HiveMQ's pricing page, as listed on 20 November 2026, says all enterprise extensions are included in every self-managed plan, so check the plan limits rather than a per-extension fee.

Does HiveMQ have an Azure Event Hubs extension?

We did not find one in HiveMQ's extension catalogue when we read it on 20 November 2026. The catalogue lists Kafka, Google Pub/Sub and Amazon Kinesis among its cloud messaging entries. Event Hubs offers a Kafka-compatible endpoint, but we have not tested the Kafka extension against it. Ask HiveMQ before you design around it.

What is HiveMQ Data Hub?

Data Hub is a feature inside the HiveMQ broker that validates and shapes MQTT data as it arrives. Data policies check payloads against JSON Schema or Protobuf schemas and can drop, log or reroute failures. Behaviour policies watch what clients do. Transformations use JavaScript. It has its own page in this cluster.

What breaks when you upgrade HiveMQ with custom extensions?

The extension has to be rebuilt and tested against the new SDK version. HiveMQ's documentation says several Enterprise SDK services and methods have been deprecated since 4.53 and will be removed in the next LTS release. A Java version change, such as the Enterprise SDK compile target rising to Java 21 in 4.55, can also force a rebuild. Plan an extension test in every upgrade rehearsal.

References and further reading

  1. HiveMQ extensions catalogue · HiveMQThe 29 listed extensions and their classes, as read on 20 November 2026.
  2. HiveMQ extension development · HiveMQCommunity and Enterprise SDK, Java requirement, the 4.53 deprecations.
  3. HiveMQ Data Hub · HiveMQPolicies, schemas, free mode and the 5-hour trial.
  4. What's new in HiveMQ 4.55 · HiveMQThe current LTS, Java 21 SDK compile target, Kafka extension validation change.
  5. HiveMQ pricing · HiveMQEnterprise extensions and in-stream data governance listed in the plans, as read on 20 November 2026.
  6. HiveMQ Amazon Kinesis extension documentation · HiveMQStates that commercial extensions ship preinstalled and disabled.

← HiveMQ

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.