---
title: "What is agent orchestration? Patterns for AI agents"
description: "Agent orchestration is the logic that decides which agent, model call, or tool runs next. Learn the main patterns, from orchestrator-workers to handoffs."
type: "glossary-term"
canonical_url: "https://theaiengineering.com/glossary/agent-orchestration"
markdown_url: "https://theaiengineering.com/glossary/agent-orchestration.md"
topic: "AI agents"
author: "The AI Engineering Editors"
publisher: "The AI Engineering"
updated: "2026-11-06"
keywords: ["Agent orchestration","multi-agent orchestration","orchestrator-worker pattern","orchestrator-workers"]
citation: "The AI Engineering Editors. \"What is agent orchestration? Patterns for AI agents.\" The AI Engineering, updated Nov 06, 2026. https://theaiengineering.com/glossary/agent-orchestration"
---

# What is agent orchestration? Patterns for AI agents

Also called multi-agent orchestration, orchestrator-worker pattern, orchestrator-workers.

> **Definition of Agent orchestration:** Agent orchestration is the control logic that decides which agent, model call, or tool runs next in an AI system and what each receives, whether written in code or chosen by a model.

Agent orchestration is the control logic that decides which agent, model call, or tool runs next in an AI system, and what each one receives as input. The logic can live in code, as a fixed pipeline or graph, or in a model, as an orchestrator agent that plans and delegates at run time. The orchestration model is the main practical difference between agent frameworks, and it determines how predictable, testable, and expensive a system is.

## How agent orchestration works

Agent orchestration works in one of two ways. In code-driven orchestration, the developer writes the sequence, branches, and parallel steps, and model output fills in the content. In model-driven orchestration, an orchestrator model reads the task and decides which worker to call next. The [OpenAI Agents SDK documentation](https://openai.github.io/openai-agents-python/multi_agent/) describes both approaches and notes that they can be mixed.

Anthropic's [Building effective agents](https://www.anthropic.com/engineering/building-effective-agents) names five workflow patterns, and [Microsoft Agent Framework](https://learn.microsoft.com/en-us/agent-framework/workflows/orchestrations/) adds multi-agent patterns of its own:

| Pattern | Who decides the next step | Example |
|---|---|---|
| Prompt chaining (sequential) | Code, in a fixed order | Outline, then draft, then edit |
| Routing | A classifier step, then code | Send a ticket to the billing or technical prompt |
| Parallelization (concurrent) | Code fans out and merges | Several reviewers score one answer |
| Orchestrator-workers | An orchestrator model | A lead agent splits research across [subagents](https://theaiengineering.com/glossary/subagent) |
| Evaluator-optimizer | Code loops on a critic's verdict | Draft, critique, and revise until a check passes |
| Handoff | The active agent | Triage passes the user to a refund agent |
| Group chat and Magentic | A shared conversation or a manager agent | Specialists coordinated by a manager |

In the orchestrator-workers pattern, Anthropic describes a central LLM that "dynamically breaks down tasks, delegates them to worker LLMs, and synthesizes their results." It differs from parallelization because the orchestrator chooses the subtasks from the input instead of following a list fixed in code.

## Orchestration vs choreography

Orchestration routes work through a central coordinator, while choreography lets each agent decide where work goes next. A coordinator sees the whole task, which makes tracing, budgets, and retries easier to manage in one place. A choreographed design removes the coordinator, but it spreads the routing logic across every agent's instructions, so a single run is harder to follow.

[Agent handoffs](https://theaiengineering.com/glossary/agent-handoff) are a choreographed design. Microsoft Agent Framework implements handoff orchestration as a mesh in which agents connect directly without an orchestrator, while its Magentic pattern uses a manager agent to coordinate specialists. Across organizations, the [A2A protocol](https://theaiengineering.com/glossary/a2a-protocol) lets an orchestrator agent delegate tasks to remote agents whose memory and tools stay hidden from it.

## Orchestration vs a single agent loop

A single agent loop has no orchestration layer: one model calls tools until it finishes. Add orchestration when a measured limit appears, such as a context window filled by exploration, independent subtasks that could run in parallel, or tools that need separate permissions. Anthropic recommends finding the simplest solution possible and adding complexity only when needed.

Each layer of orchestration adds model calls. Anthropic's [multi-agent research system](https://www.anthropic.com/engineering/multi-agent-research-system) used about 15 times the tokens of a chat interaction, so a multi-agent design has to buy enough quality to cover that cost. [Multi-agent systems](https://theaiengineering.com/glossary/multi-agent-system) also need evaluation of the whole trajectory, since a correct final answer can hide a wasteful path.

The [AI agent frameworks comparison](https://theaiengineering.com/articles/ai-agent-frameworks) sorts current frameworks by orchestration model, from plain tool loops to graphs, handoffs, and role-based crews.

## Related terms

- [Multi-agent system](https://theaiengineering.com/glossary/multi-agent-system): A multi-agent system is an AI architecture in which several language-model agents, each with its own prompt, tools, and context, work on parts of a task and pass results to one another.
- [Subagent](https://theaiengineering.com/glossary/subagent): A subagent is an AI agent that another agent starts for one subtask, running in a separate context window with its own instructions and tools, and returning only a condensed result.
- [Agent handoff](https://theaiengineering.com/glossary/agent-handoff): An agent handoff is a transfer of control in which one AI agent passes the conversation, and responsibility for the task, to another agent better suited to the next part of the work.
- [AI agent](https://theaiengineering.com/glossary/ai-agent): An AI agent is a software system in which a large language model runs in a loop, calling tools, reading their results, and repeating until the task is complete or a stop condition is reached.
- [Agentic AI](https://theaiengineering.com/glossary/agentic-ai): Agentic AI is the class of AI systems that pursue a goal through multiple steps, using a language model to plan, call tools, and act on the results with limited human input.
- [A2A protocol](https://theaiengineering.com/glossary/a2a-protocol): The A2A (Agent2Agent) protocol is an open standard, introduced by Google in April 2025, that lets AI agents built on different frameworks discover each other, exchange messages, and delegate tasks.

## Guides that cover agent orchestration

- [AI agent frameworks compared: LangGraph, CrewAI, ADK, and more](https://theaiengineering.com/articles/ai-agent-frameworks): An AI agent frameworks comparison of LangGraph, LangChain, CrewAI, the OpenAI, Claude, and Google SDKs, Pydantic AI, Mastra, and more, by language and license.
- [What are AI agents? How agents work and when to use one](https://theaiengineering.com/articles/what-are-ai-agents): AI agents are LLM systems that call tools in a loop until a task is done. Learn how the loop works, agents vs agentic AI, examples, failure modes, and limits.
- [A2A protocol vs MCP: differences and when to use each](https://theaiengineering.com/articles/a2a-vs-mcp): A2A protocol vs MCP: A2A connects AI agents to other agents, and MCP connects an agent to tools and data. Compare agent cards, tasks, transports, and security.
- [How to build an AI agent: a step-by-step guide in Python](https://theaiengineering.com/articles/how-to-build-an-ai-agent): Learn how to build an AI agent from scratch in Python: when you need one, tool schemas, the agent loop, stop conditions, guardrails, evals, and tracing.
- [AI agent evaluation: metrics, graders, and how to test agents](https://theaiengineering.com/articles/ai-agent-evaluation): AI agent evaluation guide: measure task success, tool use, safety, cost, and pass^k reliability with code graders, LLM judges, eval sets from traces, and CI.
- [AI engineer roadmap: skills, projects, and resources by stage](https://theaiengineering.com/articles/ai-engineer-roadmap): A practical AI engineer roadmap in seven stages, each with the skills to learn, a project to ship, and resources worth the time, from Python to production.
- [Agentic workflows: five patterns and when to use each](https://theaiengineering.com/articles/agentic-workflows): Agentic workflows run LLM and tool calls through steps fixed in code. Learn five patterns, their cost and latency, failure modes, and when to use an agent.

Part of the [AI agents](https://theaiengineering.com/topics/ai-agents) topic hub.

## Cite this definition

The AI Engineering Editors. "What is agent orchestration? Patterns for AI agents." The AI Engineering, updated Nov 06, 2026. https://theaiengineering.com/glossary/agent-orchestration

- Title: What is agent orchestration? Patterns for AI agents
- Author: The AI Engineering Editors
- Publisher: The AI Engineering
- Updated: 2026-11-06
- URL: https://theaiengineering.com/glossary/agent-orchestration
- Markdown: https://theaiengineering.com/glossary/agent-orchestration.md
