Coreflux MQTT Broker bridges IT and OT: native Connectors read PLCs, and your logic shapes the data before it reaches your databases and IT systems.
Bridges join brokers. Clusters make one broker bigger. Big sites do both.
Ideal for:
Testing on one line, one machine cell or one station.
Connectors read the machines, Actions react, Models store readings in a database and a Dashboard shows them. All of it runs on one machine, even a small edge device.
Available in every Coreflux plan.
The whole pipeline in one process.
The site broker keeps acquisition and control local. A bridge forwards chosen topics to a cloud broker over one outbound TLS connection, so no inbound port opens on the plant network. If the link drops, the site keeps running.
Ideal for:
Corporate wants dashboards and IT wants the firewall shut. Both get their way.
Available in the Coreflux Growth plan and up.
Control stays on site. Data goes up.
Each line or site runs its own broker and forwards only clean, named topics to a site or enterprise broker. There, Models build one namespace from machine tags up to KPIs, and Rules decide who sees what. A cut WAN link slows reporting, never the plant.
Ideal for:
You run many lines or many sites, or a plant network you don't trust.
Available in the Coreflux Growth plan and up.
Every line and every site in one namespace.
The nodes share the load. When the leader fails, the others elect a new one, typically in a fraction of a second. QoS 1 and 2 publishes are copied across nodes, so losing a minority of the cluster can't erase a message, and clients of a failed node reconnect to a healthy one.
Ideal for:
The broker can't go down, or you have tens of thousands of clients.
Available in the Coreflux Enterprise plan.
Three or five brokers acting as one.
One broker
The whole pipeline runs in a single process, even on a small edge device.
Dual broker
Send only the topics you choose up to the cloud (opens in a new tab), over one outbound connection.
Cluster
Three or five nodes act as one broker (opens in a new tab) and share the load. Need more headroom? Add a node and you got it.
Fully MQTT 3.1.1 and 5.0 compliant.
QoS 0, 1 and 2, retained messages, persistent sessions and last will, over TCP, TLS and WebSockets.
Turn on durability, and retained messages, sessions and queued messages survive a restart.
Config, users and LoT are text files, so the whole Broker lives in Git and ships through the pipeline you already run.
Health and metrics endpoints (opens in a new tab), system topics (opens in a new tab) and live logs show what it's doing, and every stored project publishes a SHA-256 manifest of exactly what's deployed.
TLS on every plan.
Rules set who may connect, publish and subscribe on each topic, in plain text you review and version like code. Enterprise adds mTLS and role-based access control.
Bridges dial out, so no inbound port opens on the plant network, and passwords stay in secrets (opens in a new tab), out of your code.
Run the Broker on your own servers and edge devices, next to the machines it listens to.
Deploy the same Broker in the cloud account you already use.
The same Broker, free for your own projects, to learn on and try ideas.
Test your connection on a live public MQTT broker.
Every message sent to the public broker is visible to all connected clients, so keep production data off it.
Drag a pill, or focus it and use the arrow keys, to pull it away. Escape lets it go.
Actions and Models run inside the Broker, on every message as it arrives.A raw reading is scaled, checked against its limit and raised as an alarm (opens in a new tab) in one place, with no separate service to subscribe, process and publish it back.
LoT is close to plain English, so the engineer who writes it, the operator who asked for it and the auditor who signs it off all work from the same file.An AI can write it too (opens in a new tab), in the syntax you review.
Native Connectors read S7, EtherNet/IP, Modbus, OPC UA and BACnet straight into the Broker, and write to your databases, Kafka and REST APIs on the way out.The protocol gateway, the flow tool and the sync scripts in between have nothing left to do.
A Model (opens in a new tab) is a few lines of LoT (opens in a new tab). Keep it in Git, ship it with the CI/CD you already run, and the same kind of machine publishes the same topics at every site.Site against site, the numbers line up without anyone mapping them by hand.
Try it in our cloud or on your own hardware, with every Growth feature for 30 days.
