Coordinator-Subagent (Orchestrator) Pattern

Core

Orchestrate multi-agent systems with coordinator-subagent patterns · Difficulty 3/5

multi-agentcoordinatororchestrationhub-and-spoke

Prerequisites

The coordinator (orchestrator) pattern uses a central agent to manage specialized subagents, maintaining control over task decomposition, routing, and result aggregation.

Architecture

Coordinator Agent
+-- Web Search Agent
+-- Document Analysis Agent
+-- Synthesis Agent
+-- Report Generation Agent

Key Principle: Hub-and-Spoke Communication

All inter-agent communication flows through the coordinator:

Subagents NEVER communicate directly with each other

Subagents operate with isolated context -- they do not inherit the coordinator's conversation history

Coordinator observes all interactions

Coordinator handles errors consistently

Coordinator decides what information each subagent receives

Dynamic Routing

The coordinator should analyze query requirements and dynamically select which subagents to invoke, rather than always routing through the full pipeline. Simple queries may only need one subagent; complex ones may need several.

Iterative Refinement

The coordinator evaluates synthesis output for gaps, re-delegates to subagents with targeted queries, and re-invokes synthesis until coverage is sufficient.

Benefits

Centralized visibility: See all interactions in one place

Consistent error handling: One place for recovery logic

Information control: Filter what each agent sees

Flexible routing: Change downstream agents without affecting upstream

Key Takeaways

✓All communication flows through the coordinator (hub-and-spoke)
✓Subagents never communicate directly with each other and have isolated context
✓Coordinator provides centralized visibility, error handling, and information control
✓Dynamically select subagents based on query complexity rather than always running the full pipeline

Glossary Terms

Coordinator/Orchestrator Pattern

A multi-agent architecture where a central coordinator manages specialized subagents using hub-and-spoke communication. The coordinator handles task decomposition, routing, and result aggregation. Subagents never communicate directly with each other, keeping the system auditable.

Related Concepts

Subagent Invocation & the Task Tool

The Task tool spawns subagents; allowedTools must include "Task" for the coordinator

Task Decomposition & Routing Strategies

Use prompt chaining for predictable workflows, dynamic decomposition for open-ended tasks

Test Yourself1 of 2

You're designing the communication pattern between your coordinator and subagents. Which architecture best supports visibility, error handling, and information control?