Swarm, graph, delegation, critique — and when each one is wrong.

Module 07 of 16 · track Build · Estimated time: 6–7 hours · Prerequisites: Module 06.

Multi-agent is not automatically better. This module teaches the patterns as a menu with costs attached: what each buys you, what it costs in tokens and latency, and the failure mode each invites.

Open the module on GitHub ↗: the slides, notebooks, exercises and solutions are there. This page is the reading order, with the tutorial and the workbench alongside.

The lesson that frames this module is Agentic PDLC for Solution Architects: Authority and Topology (7 min): how many agents, and why each hand-off costs. Then try the NFR utility tree in the workbench: the same decision, with your numbers.

• Implement swarm, graph, delegation and critique patterns in Strands
• Choose a pattern from requirements rather than fashion
• Predict the token and latency cost of a topology before running it
• Debug a multi-agent system where the failure is in the handoff, not the agent

Portable GenAI concepts	AWS-specific surface
Swarm	Strands multi-agent primitives
Graph orchestration	Bedrock model routing
Delegation
Critique/reflection loops
Deterministic vs autonomous orchestration

Portable concepts transfer to any stack. The AWS column is where this module touches the cloud — see docs/concepts/portability-matrix.md.

#		Step	What it is
1	📖	slides/Advanced_Strands_Multi_Agent_Patterns_SLIDES.md	Advanced multi-agent patterns
2	💻	notebooks/NB1_Foundations_Workflow_Patterns.ipynb	NB1 — workflow patterns
3	✏️	exercises/Exercise_1_Foundations.md	Exercise 1 — foundations
4	✏️	exercises/Exercise_2_Dials_and_Cost.md	Exercise 2 — dials and cost
5	💻	notebooks/NB2_Agentic_Patterns.ipynb	NB2 — agentic patterns
6	✏️	exercises/Exercise_3_Delegation_and_Critique.md	Exercise 3 — delegation and critique
7	💻	notebooks/NB3_Autonomous_Deterministic_Orchestration.ipynb	NB3 — autonomous vs deterministic
8	✏️	exercises/Exercise_4_Swarm_and_Graph.md	Exercise 4 — swarm and graph
9	💻	notebooks/PierPoint_Release_Desk_Graph_vs_Swarm.ipynb	PierPoint: graph vs swarm, head to head
10	✏️	exercises/Exercise_5_Capstone_Composition.md	Exercise 5 — capstone composition
11	✏️	exercises/Coding_Exercise_3_Advanced.ipynb	Advanced coding exercise

📖 read   💻 run   ✏️ practise   📊 workbook   🔖 reference

Worked solutions live in solutions/. Attempt every exercise closed-book first — the solutions are written to be read after you have a wrong answer to compare against.

• Adding agents to fix a prompt problem. More agents multiply the prompt problem.
• Swarms with no termination condition. They will happily run until your budget ends.
• Ignoring that every handoff re-sends context. Topology is a cost decision.

Every error anyone has hit in this course is in the Error Index, by the exact string. Model access and region problems are in Model and Region Notes. If it is a question rather than an error, check the answered ones first, then ask in Q&A.

Frameworks

• Handoff Multiplier — H× — what a topology costs versus one agent
• Cost Cliff Map — Unbounded swarms are cliff 3
• Three Clocks — Turns multiply two of the three clocks

Quick reference

• Strands — Graph, swarm, delegation in code

Recipes and procedures

• Choose a topology — Choose one and justify it in an hour

---

← Module 06 · Strands Foundations: Tools, Memory and MCP · All modules · Module 08 · LangChain and LangGraph →

_{✏️ This page is generated from modules/07-strands-multi-agent-patterns/README.md. To change it, edit the module README: an edit made here is replaced at the next sync.}