MCP Mastery
About
Chapter 10
nightmare
~90 min

Capstone: Multi-Agent Triage

LangGraph supervisor, two Runtime workers, Memory, Identity, Gateway MCP tool.

LangGraph ^0.4
Python >=3.11
runtime
memory
identity
gateway
observability
Reviewed 2026-05-16

Reading this chapter helps prevent 4 common multi-agent mistakes.

Overview

LangGraph supervisor, two Runtime workers, Memory, Identity, Gateway MCP tool.

Capstone: Multi-Agent Triage architecture.

Key ideas

  • LangGraph owns orchestration: explicit state, nodes, and conditional edges.
  • AgentCore owns production concerns: runtime hosting, memory, identity, gateway, observability.
  • MCP standardizes tool surfaces so workers do not hard-code every backend integration.

Labs 10 in labs/agentcore/ follow a full deploy progression. Read the lab README for IAM and cost notes.

Walkthrough

  1. Model the workflow as a graph: who plans, who executes tools, who summarizes.
  2. Attach MCP tools through Gateway (or local mocks in early labs).
  3. Add Memory for session continuity and Identity before calling protected APIs.
  4. Instrument traces and run golden-set evals before promoting changes.

References

Quiz

  1. What is the primary focus of Capstone: Multi-Agent Triage?

  2. Which pattern routes work between specialized agents?

  3. Why expose tools through AgentCore Gateway?