Edwy
AI Primer

Agentic AI · Module 21

Multi-agent systems

Once you have one agent, the next step is many. They specialise, coordinate, escalate, and audit each other.

Orchestrator

Coder

Reviewer

Deployer

Auditor

Researcher

Planner

Alfred is a multi-agent operator OS — a substrate that schedules, observes, and audits work performed by other agents.

Applied frame

Convert capability into a bounded work loop with tools, memory, permits, review, and audit.

Diagram cue

The diagram keeps autonomy wrapped in observable control points.

Multi-agent system diagram with orchestrator routing work to research, builder, and reviewer agents
Multi-agent systems need coordination as much as specialization.
Multi-agent systems animated explainer
Owned Edwy animated explainer for Multi-agent systems.

Multiple agents need routing, specialization, and escalation. Key motion beats: orchestrator, research, builder, reviewer.

Multi-agent orchestration demo

A local toy loop that logs coordination steps for specialized agents, matching this module’s orchestration framing.

const agents = ['research', 'implementation', 'review'];
for (const agent of agents) {
  console.log('coordinate agent:', agent);
}
Run the demo to see sandboxed console output.

Multi-agent systems check

0 of 1 questions completed locally.

1. What does an orchestrator help multiple agents do?

Answer feedback appears here.

Reader progress is stored locally in this browser.

Source slide 22