MODULE 7 · DAY 2
The Multi-Agent Incident Crew
One agent grows into a crew of four — sharing one model
Gourav Shah · School of DevOps & AI · Hands-on
M7·01
What you'll learn
Why a crew, when one agent is enough, and the two ways to build one.
M7·02
One agent doing everything gets unfocused
Cram triage, retrieval, fixing and safety into one prompt and roles blur.
M7·03
A hospital, not a superhero
No single doctor greets, diagnoses, prescribes, and signs their own order.
M7·04
Three things a single agent can't replicate
The properties that earn the extra prompts and latency.
M7·05
When is a single agent enough?
Multi-agent adds real overhead. Don't reach for it by default.
M7·06
The Incident Crew pipeline
A sequential pipeline with a relevance gate that can short-circuit.
M7·07
Two paths to multi-agent
Declarative profiles by default; a framework when you need more control.
M7·08
Swap the orchestrator, not the tools
Skills, MCP tools, and guardrails are the same layer under every crew.
M7·09
One model, four agents
Agents are cheap Python calls. The model is expensive — and shared.
M7·10
Wire the crew with the Compose Spec
Each agent a lightweight service, all sharing one endpoint and memory.
M7·11
The Reviewer — human-in-the-loop proxy
It doesn't decide to fix; it decides if the fix is safe to show a human.
M7·12
TO THE LAB
Swap the orchestrator, keep the crew
Read the four profiles, run both incidents, watch the gate work.
Then Module 8 — harden & ship. · Gourav Shah · School of DevOps & AI
M7·13