Mozaik in Plain English: A Gentle Introduction to Concurrent AI Agents
Most AI agent tutorials show you a pipeline. Agent A extracts data, hands it to Agent B, B hands it to C. It works, but it works like an assembly line: if A is slow, everyone waits. If B fails, the line stops. And if you want to add a new specialist to the line, you have to rewire the whole thing. Mozaik , an open-source runtime built by the JigJoy team, starts from a different idea: what if…
Mozaik is an open-source runtime that reimagines how AI agents collaborate. Instead of agents following a strict pipeline, Mozaik treats agents as participants in a shared "room" that everyone joins. When an event occurs, like someone sending a message or adding telemetry data, it becomes an event on a shared bus. Each participant can then observe the events and decide how to react independently.
Key attributes of Mozaik include concurrency, awareness, and adaptivity. Concurrency allows agents to work simultaneously without blocking each other. Awareness means agents are aware of other participants in the room and can react to them joining or leaving. Adaptivity enables agents to change their behavior based on the current situation.
The author built a war room called OpsRoom on Mozaik v4 to simulate a production incident response. OpsRoom features eight AI agents and four telemetry feeds working together. Agents handle tasks like proposing fixes, digging through logs, assessing risk, and writing incident reports. They react to each other's actions and events in real-time.
The author learned three important lessons from building OpsRoom. First, writing reactions to events can initially feel cumbersome but becomes intuitive once you shift your mindset. Second, using contract tokens in protocol messages helps ensure LLMs produce consistent outputs that can be reliably interpreted. Third, interception hooks let you inspect and modify the model's actions before they are executed, enabling you to prevent unsafe state changes.
Mozaik is particularly useful when multiple AI agents need to work together on a problem where the order of their operations is not predetermined. This includes research teams, incident response, code review, and data pipelines with judgment calls. If agents rarely need to interact, a traditional pipeline approach may be simpler.
You can find more information on the Mozaik website, documentation, GitHub repository, and the author's OpsRoom project for a live demo. The author invites readers who have built agents before to share their experiences with coordination approaches - whether they used pipelines, graphs, or the "room" model like Mozaik.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
