[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"glossary-orchestrator-worker::en":3,"gloss-cluster-orchestrator-worker::en":26,"gloss-next-orchestrator-worker::en":9},{"slug":4,"category":5,"name":6,"definition":7,"meta_desc":8,"faq":9,"schema_markup":9,"related":10},"orchestrator-worker","agents","Orchestrator-Worker Pattern","The orchestrator-worker pattern is a multi-agent design where one lead agent (the orchestrator) breaks a task into sub-tasks and delegates each to a specialized worker agent, then collects and synthesizes their results. Instead of a single agent juggling everything in one bloated context, the orchestrator plans and routes while each worker gets a clean, focused prompt and its own tools — a researcher agent, a coding agent, a reviewer agent. This mirrors how a team lead assigns work rather than doing it all personally. For builders, the pattern helps when a task has distinct sub-skills or when a single context window can't hold everything; workers can also run in parallel, cutting latency. The trade-offs are real: more model calls mean higher cost, and the orchestrator must merge sometimes-conflicting worker outputs. Start with a single agent, and only split into orchestrator-worker when one agent visibly struggles to keep separate concerns straight.","The orchestrator-worker pattern has a lead agent split a task among specialised workers and synthesise their results — each context stays small and focused.",null,[11,14,17,20,23],{"slug":12,"name":13},"agent","Agent",{"slug":15,"name":16},"agent-handoff","Agent Handoff",{"slug":18,"name":19},"agent-loop","Agent Loop",{"slug":21,"name":22},"autonomous-agent","Autonomous Agent",{"slug":24,"name":25},"model-router","Model Router",[27,31,35,36,37,40,43,46,49,52,55,58],{"slug":28,"category":5,"name":29,"updated_at":30},"agent-budget","Agent Budget","2026-08-24T02:46:37+00:00",{"slug":32,"category":5,"name":33,"updated_at":34},"agent-checkpointing","Agent Checkpointing","2026-08-24T02:46:38+00:00",{"slug":15,"category":5,"name":16,"updated_at":30},{"slug":18,"category":5,"name":19,"updated_at":30},{"slug":38,"category":5,"name":39,"updated_at":30},"agent-memory","Agent Memory",{"slug":41,"category":5,"name":42,"updated_at":34},"agent-sandbox","Agent Sandbox",{"slug":44,"category":5,"name":45,"updated_at":34},"agent-trajectory","Agent Trajectory",{"slug":47,"category":5,"name":48,"updated_at":30},"agentic-rag","Agentic RAG",{"slug":50,"category":5,"name":51,"updated_at":34},"computer-use","Computer Use",{"slug":53,"category":5,"name":54,"updated_at":30},"context-engineering","Context Engineering",{"slug":56,"category":5,"name":57,"updated_at":34},"dry-run","Dry Run",{"slug":59,"category":5,"name":60,"updated_at":61},"escalation-path","Escalation Path","2026-08-24T03:30:02+00:00"]