---
title: "What is an agent handoff in multi-agent systems?"
description: "An agent handoff transfers control of a conversation from one AI agent to another, often a specialist. See how handoffs work and how they differ from subagents."
type: "glossary-term"
canonical_url: "https://theaiengineering.com/glossary/agent-handoff"
markdown_url: "https://theaiengineering.com/glossary/agent-handoff.md"
topic: "AI agents"
author: "The AI Engineering Editors"
publisher: "The AI Engineering"
updated: "2026-11-06"
keywords: ["Agent handoff","handoff","agent handoffs","handoff orchestration"]
citation: "The AI Engineering Editors. \"What is an agent handoff in multi-agent systems?\" The AI Engineering, updated Nov 06, 2026. https://theaiengineering.com/glossary/agent-handoff"
---

# What is an agent handoff in multi-agent systems?

Also called handoff, agent handoffs, handoff orchestration.

> **Definition of 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.

An agent handoff is a transfer of control from one [AI agent](https://theaiengineering.com/glossary/ai-agent) to another during a run. A triage agent that recognizes a refund request hands the conversation to a refund agent, which then replies to the user and calls its own tools. The receiving agent takes ownership of the task, and the first agent stops acting. Handoffs let each agent keep a short prompt and a narrow tool set while the system as a whole covers many kinds of request.

## How agent handoffs work

Agent frameworks typically implement a handoff as a [tool call](https://theaiengineering.com/glossary/tool-calling). The framework gives the current agent one tool per allowed target, the model calls the tool for the agent that fits the request, and the runtime switches the active agent. The new agent continues from the conversation history, which the developer can filter first.

The [OpenAI Agents SDK](https://openai.github.io/openai-agents-python/handoffs/) presents each handoff to the model as a tool, so a handoff to a Refund Agent appears as `transfer_to_refund_agent`. By default the receiving agent sees the entire previous conversation. An `input_filter` function can trim that history, and an `input_type` schema lets the model attach structured data, such as a reason, to the handoff call.

[Microsoft Agent Framework](https://learn.microsoft.com/en-us/agent-framework/workflows/orchestrations/handoff) builds handoff orchestration as a mesh: agents connect directly, with no central orchestrator, and the developer declares which agents may hand off to which. Its runtime strips handoff tool calls and other tool content from the history before forwarding it, so the next agent sees user and agent messages. Handoff workflows there are interactive by default: when an agent replies without handing off, control returns to the user.

## Agent handoff vs agent as a tool

In a handoff, control moves to the receiving agent and stays there. When an agent is used as a tool, or as a [subagent](https://theaiengineering.com/glossary/subagent), the calling agent sends a subtask, gets a result back, and keeps control of the conversation. Microsoft's documentation adds task ownership to the distinction, since the agent that receives a handoff fully owns the task.

| | Handoff | Agent as a tool (subagent) |
|---|---|---|
| Who replies to the user next | The receiving agent | The calling agent |
| Context passed | The conversation history, optionally filtered | A task description chosen by the caller |
| Who owns the task | The receiving agent | The calling agent |
| Central coordinator | Not required | The calling agent |
| Good fit | Triage to specialists, such as billing or support | Bounded subtasks, parallel research |

The two combine. The [OpenAI Agents SDK](https://openai.github.io/openai-agents-python/multi_agent/) describes a triage agent that hands off to a specialist, which in turn calls other agents as tools for subtasks that should not take over the user-facing conversation.

Two design checks follow from this. The triage agent can route to the wrong specialist, so routing accuracy belongs in the [evaluation](https://theaiengineering.com/glossary/llm-evaluation) set. A long history also travels with every handoff, so pass the next agent what it needs instead of forwarding everything.

The [AI agent frameworks comparison](https://theaiengineering.com/articles/ai-agent-frameworks) shows which frameworks are built around handoffs and which use graphs, crews, or a single tool loop.

## 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 orchestration](https://theaiengineering.com/glossary/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.
- [Tool calling](https://theaiengineering.com/glossary/tool-calling): Tool calling is an LLM capability in which the model, given a list of tool definitions, outputs a structured request naming a tool and its arguments, which the application executes and returns.
- [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.

## Guides that cover agent handoff

- [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.
- [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 an agent handoff in multi-agent systems?" The AI Engineering, updated Nov 06, 2026. https://theaiengineering.com/glossary/agent-handoff

- Title: What is an agent handoff in multi-agent systems?
- Author: The AI Engineering Editors
- Publisher: The AI Engineering
- Updated: 2026-11-06
- URL: https://theaiengineering.com/glossary/agent-handoff
- Markdown: https://theaiengineering.com/glossary/agent-handoff.md
