Introducing BunkerM Enterprise: Native Sparkplug B and AI for Industrial IoT
BunkerM started as a management layer for Eclipse Mosquitto. The Community edition is built around plain MQTT and JSON payloads, which works well for home automation, small deployments, and developer tooling. For industrial environments, that model has limits. Today we are announcing BunkerM Enterprise: a new edition of the platform built around the Sparkplug B specification, where every device self-describes on connect and the AI understands your plant from the first message.
Why Sparkplug B Changes the AI Story
The core problem with AI querying plain MQTT is context. When the broker receives a message on factory/zone1/pump01 with payload {"v": 1420}, the AI has to guess what v means, what units it is in, and whether 1420 is normal. For a handful of devices that is manageable. For a production line with 200 edge nodes across multiple PLCs and SCADA systems, it breaks down fast.
Sparkplug B solves this at the protocol level. When an edge node connects, it publishes a birth certificate: a structured message that declares every metric it exposes, its name, data type, engineering unit, and current value. Pump01.Speed is a FLOAT in RPM, currently 1420.0. The AI does not need to guess. It reads the birth certificate and knows the full device model before the first query arrives.
That is the foundation BunkerM Enterprise is built on: the world's first open-source, AI-powered MQTT broker with native Sparkplug B support.
What the Platform Does
BunkerM Enterprise runs as a single Docker container, extending the Community edition with a new sparkplug-api service that subscribes to spBv1.0/#, decodes every protobuf payload, and maintains a live registry of your entire device hierarchy: groups, edge nodes, devices, and metrics, all with current values, engineering units, and quality flags.
The AI assistant connects to that registry. Instead of searching raw topic trees, it navigates a structured model. Ask it which devices are offline and it checks NDEATH/DDEATH events. Ask for the current pressure on Line 2 and it returns the metric value with the unit from the birth certificate. Ask it to set a pump setpoint and it encodes the DCMD, resolves the correct metric name from the registry, and publishes the command, with a confirmation step before anything fires.
Plain MQTT devices and Sparkplug B devices run on the same broker simultaneously. Legacy equipment does not need to be migrated or replaced.
Who It Is Built For
The target is industrial OT teams and system integrators already working in the Sparkplug B ecosystem. If you are running Siemens S7 PLCs bridged via Cirrus Link MQTT Transmission, Opto 22 groov EPIC edge nodes, Inductive Automation Ignition as your SCADA platform, or Rockwell Allen-Bradley controllers publishing through FactoryTalk Edge, BunkerM Enterprise connects directly to your existing architecture without additional middleware.
The same applies to other hardware in the field: Mitsubishi MELSEC, Schneider Electric Modicon, ABB AC500, Omron NX/NJ, and Beckhoff TwinCAT all have MQTT or Sparkplug B paths available. If the device can publish spBv1.0/#, BunkerM Enterprise picks it up.
Three Use Cases Worth Naming
Fault isolation without the SCADA client
A production line loses throughput mid-shift. The supervisor pulls up the AI chat on a phone and asks: "Which devices on Line 3 reported bad quality data in the last hour?" BunkerM checks quality flags across all registered devices, cross-references DDEATH events, and returns the two sensors that went offline and when. The SCADA historian would have given the same answer in about ten minutes, after logging in, filtering by time range, and reading through a list of alarm states. BunkerM returns it in a sentence.
Command execution from anywhere
A grid frequency event requires adjusting pitch setpoints on six wind turbines. The operator is not at a workstation. They open Telegram, type: "Reduce pitch angle setpoint on Turbines 7 through 12 to 2.5 degrees." BunkerM resolves the six devices in the registry, identifies the correct metric name from each birth certificate, encodes a DCMD for each, and publishes them with a single confirmation prompt. No remote desktop. No SCADA client login.
Shift handover in one message
At shift change, the incoming supervisor asks: "Summarize the last 8 hours: devices that went offline, quality alerts, and lines running below target cycle time." BunkerM pulls device lifecycle events, quality flag history, and metric data and returns a plain-English summary. A task that previously meant assembling information from multiple screens across a Wonderware or AVEVA System Platform deployment.
Current Status and Early Access
BunkerM Enterprise is in active development. The Sparkplug B subscriber, protobuf decoder, and device registry are built. The REST API for AI tool calls and the enterprise AI tools on the BunkerAI Cloud side are in progress.
We are looking for a small number of design partners from industrial automation to work with us on the integration story before general availability. If you are running Sparkplug B infrastructure, evaluating open broker alternatives to proprietary SCADA stacks, or building IIoT platforms on top of Ignition or Cirrus Link, we want to hear about your setup.
Reach out at support@bunkerai.dev. The Community edition remains fully open source at github.com/bunkeriot/BunkerM.