Understand what’s happening
Stream device state, sensor readings, and events to agents, so decisions start with what is happening in the physical world.
Artificial intelligence
Connect the physical
world
to AI agents
Give AI agents live device context and a way to act. Route telemetry, events, and commands between agents and devices at scale with EMQX.
Overview
Give agents context. Give devices intelligence.
Give agents live device data to make informed decisions, act through authorized commands, and adapt to the results.
Decide what to do next
Agents combine live observations with goals, history, and device capabilities to interpret a situation and choose an action.
Turn intent into action
Deliver authorized commands to the right devices. Return execution results and updated state so agents can decide what comes next.
Architecture
Live data to agents. Actions to devices.
Connect devices with MQTT, expose device tools through MCP or Device Agent, and coordinate agents with A2A over MQTT.
Device state, events, and telemetry
Physical devices
- Robots and machines
- Sensors and equipment
- Connected products
EMQX
Connectivity and coordination
- MQTT messaging
- Routing and access control
- MCP and A2A over MQTT
AI agents
- Perceive device context
- Reason with models and tools
- Choose and coordinate actions
Tool calls and device commands
Capabilities
Connect device fleets. Coordinate AI agents.
Connect device fleets
Use clustered MQTT messaging to connect distributed devices and route their updates to the agents that need them.
EMQX clusteringMake telemetry useful
Give agents consistent device data by transforming messages before routing them.
Message transformationExpose device tools
Use MCP over MQTT to make device capabilities available as discoverable tools, with requests and results routed through EMQX.
MCP over MQTTDefine what devices can do
With Device Agent, describe commands, properties, and events so agents can select actions your devices actually support.
Define a Device AgentCoordinate across agents
Let specialist agents discover one another and exchange tasks using A2A over MQTT across distributed environments.
A2A over MQTTControl who can act
Scope each agent’s access to device data and command topics.
Access controlUse cases
Put agent intelligence to work in the real world
Coordinate robot fleets
Combine robot status with task context. Let agents coordinate assignments and send supported commands, with execution results closing the loop.
Explore connected robotsAssist factory operators
Bring equipment readings and fault events into agent workflows for diagnosis, maintenance planning, and operator-approved adjustments.
Explore smart manufacturingAdapt energy operations
Use consumption, equipment state, and operating goals to help agents recommend or request permitted changes to device settings.
Explore smart energyEnable conversational products
Turn natural-language requests into device interactions. Give agents the reported state and supported actions they need to respond usefully.
Explore connected productsCustomer story
FoloToy connects AI with everyday play
FoloToy uses EMQX to connect its AI toys with backend services for real-time interactions across regions. MQTT messaging supports the connection between physical products and the AI experiences behind them.
Read the FoloToy storyResources
Explore EMQX for artificial intelligence
FAQ
Frequently asked questions
Devices publish state and events through MQTT, directly or through a gateway. Agents consume that context and invoke device capabilities through an MCP server, Device Agent, or an application integration. EMQX routes the messages between them, including commands and results.
The agent’s model and runtime interpret context and choose actions. EMQX provides the messaging and connectivity layer. You can use your chosen agent framework, or build device interactions with EMQX Agents.
An existing device can keep its MQTT implementation. An MCP server or application adapter can expose its supported operations as tools and translate requests into the device’s command format. MCP over MQTT carries those tool interactions; devices still implement and report their own actions.
Use EMQX authentication and authorization to scope access to MQTT topics and services. Define permitted operations and parameter limits in the device or tool layer, and add approval steps in your application where needed. Devices should report execution results so agents can distinguish a requested action from a completed one.
EMQX clustering distributes MQTT connections and message routing. MCP over MQTT supports discovery and multiple server instances, while A2A over MQTT supports agent discovery and task exchange. Design topic permissions, device identities, and agent responsibilities around your fleet and deployment.
Bring AI agents into the physical world
Connect your devices. Give agents context. Turn decisions into action—with EMQX.