Orchestrator–Workers
A central orchestrator agent plans and delegates to specialised worker agents (or tools), then composes their outputs.
This page is the complete instruction page for one pattern called "Orchestrator–Workers." It explains the problem this pattern solves, the idea behind the solution, when you should (and should not) use it, and what happens afterward — both the good effects and the costs.
This matters because building AI agents is not just about making them clever. It is also about making them safe and predictable. Following a well-tested pattern like this one helps avoid common mistakes, and shows you exactly which safety rules and regulations it connects to, listed under "Standards Mesh" on this page.
Context
Complex tasks exceed any single agent's reliable scope. Delegation to specialists — each with narrow responsibility, its own tools, its own prompt — scales better than one generalist.
Problem
Single-agent systems do not scale in capability past a point; flat multi-agent swarms lack coordination.
Forces
- Centralised control vs worker autonomy
- Orchestrator as bottleneck vs decentralisation
Solution
Orchestrator holds the plan and the goal state. It routes sub-tasks to workers (which may themselves be agents or tools). Workers return structured results; orchestrator composes, critiques, and decides whether to iterate. Routing can be LLM-driven, deterministic, or hybrid.
Applicability
- Research agents
- Complex customer workflows
- Enterprise automation spanning multiple domains
Anti-Patterns
- Workers with overlapping responsibility
- Orchestrator that micromanages every step
Consequences
- +Scales past single-agent limits
- +Clear accountability boundaries
- −Coordination overhead
- −Orchestrator bottleneck risk