orchestrator–worker pattern
In plain wordsA way of organizing AI teamwork, like a newsroom: an editor AI decides which stories a big assignment needs, sends reporter AIs to cover them at the same time, and turns what they file into one report.
An agent architecture with a lead model, the orchestrator, that reads the task, works out at run time which subtasks it calls for and hands each to a worker — a separate model call or subagentsubagent, n.: A helper AI agent that a main agent starts to handle one piece of a job; it works in its own separate workspace and reports back a short summary when it is done. with its own instructions, tools and context windowcontext window, n.: How much an AI model can read and keep in view at once: the instructions, the conversation so far and any attached documents. Text that does not fit has to be cut, summarized or left out., often running in parallel with the others — then merges what comes back, sending out further rounds if gaps remain. As framed by Anthropic's 2024 guide that named the pattern, the difference from prompt chainingprompt chaining, n.: A way of getting a big job done with a relay of AI requests: each handles one stage and passes its result to the next — pull out the facts, write, polish — with automatic checks between stages. or fixed parallel 'sectioning' is flexibility: subtasks come from the orchestrator's reading of each input, not from code. The pattern suits breadth-first problems such as research or searching a large codebase. Anthropic reported in June 2025 that a research system with a Claude Opus 4 lead and Claude Sonnet 4 workers beat a single Opus 4 agent by 90.2% on an internal evaluation while using about 15 times the tokens of a chat. Weaknesses are coordination overhead, duplicated work and poor fit for tightly coupled tasks, including much of coding.
We moved literature search to an orchestrator–worker pattern: the lead agent spawns one worker per subtopic.
Origin
The manager–worker (master–worker) pattern is long established in parallel computing. For language models, 'orchestrator-workers' was named as one of five workflow patterns in Anthropic's 'Building effective agents' (Erik Schluntz and Barry Zhang, December 2024); Anthropic's June 2025 account of its multi-agent research system made the lead-agent-and-subagents design widely copied.
Related
- orchestrationorchestration, n.: Coordinating several AI calls, tools or agents so each does its part in the right order and their results are combined into one finished piece of work, as a conductor directs an orchestra.
- subagentsubagent, n.: A helper AI agent that a main agent starts to handle one piece of a job; it works in its own separate workspace and reports back a short summary when it is done.
- multi-agent systemmulti-agent system, n.: A setup in which several AI agents, each with its own role or point of view, work together — or compete — on a problem, passing messages and results among themselves.
- task decompositiontask decomposition, n.: Breaking a big, vague job into smaller, concrete steps that can each be done and checked — the way a recipe splits a meal into stages.
- deep researchdeep research, n.: An AI feature that spends minutes rather than seconds on a question — planning, running dozens of web searches, reading many sources — and then writes a long report with citations.
- agentic workflowagentic workflow, n.: A multi-step process in which AI models draft, check, use tools and revise their work over several rounds, instead of answering in one shot; often the steps are laid out by developers in advance.
- prompt chainingprompt chaining, n.: A way of getting a big job done with a relay of AI requests: each handles one stage and passes its result to the next — pull out the facts, write, polish — with automatic checks between stages.
Sources
- Schluntz, E. & Zhang, B. (2024). Building effective agents. Anthropic, 19 December 2024. anthropic.com/research/building-effective-agents
- Anthropic (2025). How we built our multi-agent research system. Anthropic Engineering, 13 June 2025. anthropic.com/engineering/multi-agent-research-system
Mentioned in
deep researchdeep research, n.: An AI feature that spends minutes rather than seconds on a question — planning, running dozens of web searches, reading many sources — and then writes a long report with citations., agentic workflowagentic workflow, n.: A multi-step process in which AI models draft, check, use tools and revise their work over several rounds, instead of answering in one shot; often the steps are laid out by developers in advance., multi-agent systemmulti-agent system, n.: A setup in which several AI agents, each with its own role or point of view, work together — or compete — on a problem, passing messages and results among themselves., orchestrationorchestration, n.: Coordinating several AI calls, tools or agents so each does its part in the right order and their results are combined into one finished piece of work, as a conductor directs an orchestra., plan-and-executeplan-and-execute, n.: An agent design that first writes out a complete step-by-step plan for a task, then carries out the steps one by one, revising the plan only when something goes wrong., subagentsubagent, n.: A helper AI agent that a main agent starts to handle one piece of a job; it works in its own separate workspace and reports back a short summary when it is done., task decompositiontask decomposition, n.: Breaking a big, vague job into smaller, concrete steps that can each be done and checked — the way a recipe splits a meal into stages., agent swarmagent swarm, n.: A large group of AI agents working at the same time on parts of a problem, usually with little central control — a name loosely borrowed from swarms of insects., factored cognitionfactored cognition, n.: The idea that a hard thinking task can be broken into many small, separately checkable pieces, each simple enough for a person or an AI to handle in a short time., swarm scalingswarm scaling, n.: Getting better AI results by putting many AI agents on a task at once instead of letting one think longer — faster, but more expensive for each gain.