Session 7· 03· 12 min

The 5 Multi-Agent Patterns

What you'll learn
  • Name and describe all 5 multi-agent patterns
  • Know when each pattern is the right fit
  • Understand the concept of context engineering across agents

Once you decide to go multi-agent, the next question is: how should the agents communicate and coordinate? There are five established patterns, each with distinct trade-offs. Let's walk through them.

1. Subagents (Supervisor pattern)

A supervisor agent orchestrates child agents. It decides which subagent to call, passes context, and synthesizes results. The subagents do not talk to each other directly — all communication goes through the supervisor.

When to use: When you need centralized control and the supervisor must make strategic decisions about which agents to invoke and in what order.

Subagents (Supervisor)
Main agent dispatches subagents as tools. All routing through the supervisor.
User
Supervisor Agent
orchestrates, holds memory
calls as tools ↓
Subagent A
research
Subagent B
writing
Subagent C
analysis
↑ results return to supervisor

2. Handoffs (Sequential transfer)

Agents pass control to each other in a conversation-like flow. Agent A handles the first part of a task and then "hands off" to Agent B with the relevant context. The conversation state flows from one agent to the next.

When to use: Conversational systems where different stages require different specializations — like triage, then specialist, then follow-up.

Handoffs
Agents transfer control to each other via tool calls based on conversation state.
Triage Agent
collects info
transfer_to →← transfer_to
Specialist Agent
resolves issue
transfer_to →
Billing Agent
payment
Each agent talks directly to the user. State persists across handoffs.

3. Skills (Tool-like agents)

Agents are exposed as tools that a parent agent can call. Each "skill" agent is a black box — the parent calls it with inputs and gets outputs back, just like calling a function. The skill agent manages its own context internally.

When to use: When sub-tasks are well-defined and self-contained. The parent does not need to see the skill agent's internal reasoning — just the result.

Skills
Single agent loads specialized prompts and context on demand.
Single Agent
maintains control
loads skill context ↓
Calendar skill
Email skill
CRM skill

4. Router (Dynamic dispatch)

A lightweight routing layer examines the input and dispatches it to the right specialized agent. Unlike the supervisor pattern, the router does minimal processing — it just classifies and routes. The selected agent handles the entire task.

When to use: When you have clearly distinct domains and inputs naturally fall into categories — customer support with billing, tech, and general queues.

Router
Classify input, dispatch to specialists in parallel, synthesize results.
User Query
Router / Classifier
stateless
dispatches in parallel ↓
Domain A
Domain B
Domain C
Synthesizer
combines results

5. Custom (Hybrid patterns)

Real-world systems often combine multiple patterns. A router might dispatch to subagent clusters. A handoff chain might invoke skill agents at certain steps. Custom patterns emerge when no single pattern fits your constraints.

When to use: When your requirements span multiple patterns, or when you need to optimize for a specific constraint (latency, cost, accuracy) that no single pattern addresses.

Custom Workflow
Bespoke graph mixing deterministic logic + agentic nodes. Embed other patterns as nodes.
Input
Deterministic step
Agent node
with tools
RAG node
Output
Full LangGraph — you design the graph.
Context engineering is the key insight
The real art of multi-agent architecture is context engineering — ensuring each agent receives precisely the information it needs and nothing more. Too little context and it cannot do its job. Too much and it gets confused, wastes tokens, and routes poorly. Every pattern above is essentially a different strategy for managing what context flows where.

Comparison matrix

 Centralized controlContext isolationParallelismConversationalComposable
Subagents
Handoffs
Skills
Router
Custom

Check your understanding

Knowledge Check
A customer support bot routes a user through triage, then to a billing specialist, then to a satisfaction survey — each handled by a different agent with its own tools. Which pattern is this?
Knowledge Check
What does 'context engineering' mean in multi-agent systems?
Up next
You know the patterns. But which one is fastest and cheapest? The next lesson puts hard numbers on performance and cost.