Skip to main content

Agent Creation Patterns

These patterns address a fundamental question: how should agents be structured?

Every agentic system starts with a decision about scope, authority, and coordination. A single agent handling everything is the simplest architecture — but it breaks down as tasks grow in complexity. When that happens, you need to decide how agents relate to each other: through fixed roles, hierarchical delegation, or peer-to-peer handoffs.

The patterns below move from simple to complex. Start with the simplest pattern that solves your problem.


Patterns

PatternDescriptionAgentsCoordination
Single AgentOne agent handles the entire task end-to-end1None
Role-Based AgentFixed behavioral roles — plan, execute, reviewFewRouter assigns by role
Hierarchical AgentTree structure with parent-child delegationManyTop-down supervision
Swarm AgentPeer agents coordinating through handoffsManyEmergent, peer-to-peer

How to Choose

Start with Single Agent if the task fits in one context window and doesn't need separated reasoning phases. Most prototypes should start here.

Move to Role-Based when you notice a single agent conflating planning with execution, or when you need independent review and critique of outputs.

Move to Hierarchical when tasks require multi-level decomposition — a manager breaks work into subtasks, each of which may need further breakdown.

Move to Swarm when no single agent can predict what expertise is needed, and you want agents to self-organize through handoffs rather than top-down control.