Orchestrator agent
An orchestrator agent is the top-level controller in a multi-agent system — it decomposes the user's request into sub-tasks, dispatches them to specialist agents, integrates their outputs, and handles cross-cutting concerns like error recovery and progress reporting. The orchestrator is the agent the user interacts with directly.
Orchestrators bear the hardest design problem in multi-agent systems: the decomposition. A bad orchestrator over-decomposes (calls 12 specialists for a task one would handle) or under-decomposes (asks one specialist to do everything). Healthy orchestration follows the pattern: identify the natural sub-task boundaries from the user's request, dispatch in parallel where the sub-tasks are independent, sequence where they depend, integrate the results coherently. The system prompt for an orchestrator is typically large because it must describe the available specialists, when to use each, and how to compose their outputs.
Related terms
- Multi-agent system
A multi-agent system uses multiple LLM-driven agents working together — typically with role specialisation (planner, executor, critic) or domain specialisation (researcher, coder, writer).
- Specialist agent
A specialist agent in a multi-agent system is scoped to a narrow domain — code review, data analysis, document drafting, web research — with a system prompt, tool set, and context window tuned to that domain.
- Agent loop
An agent loop is the orchestration pattern where an LLM iteratively reasons, calls tools, observes results, and continues — until a terminal condition is met (task complete, max iterations reached, error).