A technical deep-dive into the A2A over MQTT specification and Skitter, an MQTT-native agent orchestration framework.
Overview
Join us for a hands-on technical session exploring how MQTT, the protocol powering billions of IoT devices, becomes the transport layer for the next generation of AI agent communication.
We'll introduce the A2A over MQTT specification, a broker-neutral standard for agent discovery and messaging that brings interoperable agent-to-agent communication to any system already running MQTT. Then we'll go deeper with Skitter, an open-source framework that puts this architecture into practice, orchestrating AI agents over MQTT with stateless supervisors, ephemeral workers, and retained-message coordination.
Whether you're building multi-agent workflows, exploring alternatives to centralized orchestrators, or looking to extend your existing MQTT infrastructure into the AI agent space, this session is for you.
What We'll Cover
The A2A over MQTT Specification
A walkthrough of the specification's core design: how agents publish discoverable Agent Cards via retained messages, the request/reply/event topic conventions, and the minimal security model, all without coupling to any specific broker implementation.
Why MQTT for Agent Communication
MQTT was built for unreliable networks, low bandwidth, and massive scale. We'll explain why these properties make it a natural fit for distributed AI agent systems and how it compares to HTTP-based alternatives like Google's A2A protocol.
Skitter: MQTT-Native Agent Orchestration
A deep dive into Skitter's architecture: how it uses MQTT v5 pub/sub for stateless coordination, retained messages for crash recovery, and ephemeral workers that can run anywhere: locally, in Docker, or on cloud platforms like Fly.io.
Live Demo: Multi-Agent Workflow in Action
We'll build and run a multi-agent workflow live, showing agent discovery, task chaining with dependencies, real-time monitoring via the web dashboard, and how the system recovers from worker failures, all coordinated through an MQTT broker.
Q&A
Open discussion with the audience.
Who Should Attend
This webinar is ideal for engineers, architects, and technical leaders working with IoT/MQTT infrastructure, AI agent systems, or distributed computing, especially those evaluating how to integrate agentic AI capabilities into existing messaging architectures.
Speakers
- Zaming Shi CTO - EMQX, EMQ Technologies
- Ivan Dyachkov Field CTO - EMEA, EMQ Technologies
Event Details
- Date: Tuesday, April 7, 2026
- Time: 8:00 AM PDT · 10:00 AM CDT · 11:00 PM EDT · 4:00 PM BST · 5:00 PM CEST
- Duration: 45 min + 15 min Q&A
About
EMQX is the world's most scalable open-source MQTT messaging platform, connecting over 250 million IoT devices across 1,000+ enterprise customers in industries including automotive, manufacturing, energy, and telecommunications. The A2A over MQTT specification extends MQTT's proven messaging model into the emerging field of AI agent interoperability.
Skitter is an open-source personal AI assistant framework that leverages MQTT's pub/sub architecture for distributed agent orchestration, replacing centralized controllers with a stateless, broker-mediated coordination model.