multi-agent-patterns

Deploy a central supervisor agent that decomposes user objectives into subtasks, routes them to specialized workers, and aggregates results. Constrain worker output schemas to prevent supervisor context saturation, and cap workers per supervisor at 3-5.

Supervisor patterns use a central coordinator that delegates to workers and synthesizes results, giving strict control but creating a bottleneck. Swarm patterns let any agent hand off directly to any other, removing single points of failure but increasing coordination complexity.

LangGraph uses graph-based state machines with explicit nodes and edges for deterministic routing. AutoGen uses conversational GroupChat patterns for event-driven coordination. CrewAI uses role-based hierarchical crews for structured team workflows.

Production data shows multi-agent systems cost approximately 15x the tokens of a single-agent chat, versus roughly 4x for a single agent with tools. Budget for coordination overhead, retries, and consensus rounds, not just per-agent costs.

LLMs have an inherent bias toward agreement, so agents converge on agreeable rather than correct answers, and simple majority voting weights hallucinations equally with sound reasoning. Counter this with confidence-weighted voting and explicit adversarial roles.

Avoid multi-agent designs when a single agent's context can hold the task, when subtasks cannot run in parallel, or when coordination overhead exceeds the work itself. Over-decomposition into many fine-grained agents spends more tokens on handoffs than on actual work.