agentsop-agent-topology-selection

Choose the right multi-agent topology by applying a binary-question rubric (Q0–Q2) to decide between single-agent, sequential, supervisor, swarm, or hierarchical structures before writing any agent code.

Use a single agent with tools when the topology selection rubric determines that routing complexity is low and a multi-agent framework would introduce unnecessary token overhead and latency.

Supervisor topologies route tasks to individual agents, while hierarchical teams nest groups of agents under supervisors to handle complex routing and auditability constraints in multi-agent design.

Map an agent topology across frameworks by translating the selected structure into practical equivalents and handoff patterns specific to CrewAI, LangGraph, and similar frameworks.

Check for framework default misalignments and enforce termination discipline for runtime handoff topologies to prevent infinite routing loops and safeguard correctness before implementing multi-agent design.

Token optimization in multi-agent design requires choosing topologies like sequential or supervisor over swarm structures when token and latency constraints matter, ensuring efficient routing.