FRENES

Glossary · Agents and autonomy

Subagent

A subagent is a component that receives part of a task from a main agent or coordinating system. It works on an assigned objective and returns a result, potentially using a separate context, a restricted set of tools and permissions defined for that assignment.

Subagent: the explanation

A coordinating system divides an objective into smaller assignments and passes each one to the appropriate subagent. The recipient may take several actions before returning an answer. One subagent might examine registration conditions while another looks up schedules. The coordinator then combines their findings. Depending on the application, these assignments run at the same time or in sequence. A subagent may use the same underlying model as the main agent or a different one for its work.

This arrangement helps separate responsibilities and avoids giving every component all the available information. A narrow assignment can be easier to follow than a long task involving unrelated goals. It also allows particular tools to be reserved for particular roles. Product terminology varies, however. Something called a subagent might make many decisions or follow a tightly defined route. Its actual capabilities and responsibilities tell you more than the label alone about how it will behave.

Delegation creates additional work in passing instructions and checking results. An incomplete assignment may lead a subagent to answer the wrong question. Its summary can also omit an important limitation. The coordinator needs to handle these problems and reconcile conflicting findings. Several subagents using similar sources can share the same mistake. Increasing their number therefore does not replace good evidence, clear ownership of each assignment or verification suited to the objective you actually want to achieve.

A concrete example

You are planning a community event and ask for information about several venues. The system assigns capacity checks to one subagent and accessibility conditions to another. Both return sources and uncertainties. The main agent combines the findings in a comparison, flags conflicting opening hours and asks whether that point should be checked further before you proceed with selecting a suitable venue for the event.

Why it matters

Subagents can help organize work, but they can also add communication overhead, delay and cost. When choosing a tool, examine how it divides permissions and checks returned results. You should be able to identify a failed assignment or a missing piece of information. Well-managed delegation matters more than having a large collection of components described as a specialized team working on your behalf.

Common misconceptions

  • Several subagents do not automatically provide independent opinions, because they may use the same model, consult the same sources and repeat a shared mistake without their agreement making the underlying claim any more accurate.
  • A subagent does not necessarily have permanent memory or a different model, because its separate identity may come only from the assignment, context, tools and permissions allocated to it for the current piece of work.

Related terms: agents and autonomy

AI agent
An AI agent is a system that uses a model to decide which steps to take toward a task.
autonomous agent
An autonomous agent is an AI agent that can pursue a goal and choose several actions without involving you at every step.
multi-agent system
A multi-agent system is a collection of agents that divide tasks or exchange information to pursue an objective.
planning (AI agents)
Planning in AI agents is the preparation of steps an agent intends to follow toward a goal.
agent memory
Agent memory is the set of mechanisms that lets an AI agent retain and retrieve information useful to its work.
human in the loop
Human in the loop is an arrangement in which a person participates at selected stages of an AI system’s work.
tool calling
Tool calling is a mechanism through which a model asks software to run a specific function, such as searching documents or checking a calendar.
computer use
Computer use is a capability that lets an AI agent interact with a graphical interface by clicking, typing or scrolling.

Where this site covers it

The term appears on 12 pages of this site. A few of them:

Same family

Go further

Frequently asked questions

Is it a different model?

Not necessarily. A subagent can use the same model as the coordinator. Its separate role may come from its instructions, isolated context, available tools or the permissions assigned to the particular task it receives.

Does it see the full conversation?

That depends on the configuration. It may receive a short assignment, selected documents or a larger conversation history. Check what information is shared, particularly when the task includes personal details or confidential business material.

Does it always run in parallel?

No. Some subagents work simultaneously, while others wait for earlier results. The arrangement depends on relationships between assignments, available resources and how the application coordinates the different stages needed to complete your request.

Neutrality. No affiliate links. No vendor pays to be reviewed, rated or ranked. Reviews are directed by François Chollet, neutral and factual, and every data point is dated and sourced. How we score