Subagent
Subagent
An AI program that handles specific tasks under the direction of a main AI.
In Simple Terms
A subagent is an AI that handles a specific task assigned by a main AI. For example, when creating a document, the main AI gives overall instructions while subagents handle tasks like translation or summarization. By dividing up the work this way, complex tasks can be processed efficiently.
Behind the Name
"Subagent" combines "sub" (meaning secondary or subordinate) with "agent" (something that acts on someone's behalf). It refers to a subordinate AI that works under a main AI overseeing the whole system, handling one specialized task at a time.
Take a Closer Look!
A subagent is an AI within a larger AI system that receives instructions from a central main AI and carries out a specific task on its behalf.
A subagent is itself a full AI with its own dedicated role and settings, working through multiple steps on the assigned task while making its own judgment calls along the way. In some implementations, a subagent is given a separate, independent workspace to carry out its sequence of steps in, returning only the final result to the main AI once the work is done. Rather than having a single AI handle everything, dividing the work by specialty allows complex processes to be handled more efficiently.
In real systems, a parent-like main AI that oversees everything works together with child-like subagents that each handle individual tasks, much like a team.
The main AI controls the overall flow and delegates work to specialized subagents as needed.
Because each agent's role and settings can be managed independently, the overall system structure stays easy to follow.
Another key feature is the ability to run multiple subagents in parallel at the same time.
For example, when building a new feature, running a research subagent that gathers necessary information alongside a design-review subagent that checks the design can shorten the overall processing time, since each one proceeds with its own research or review independently. This kind of setup, where role-divided AIs work together, is widely used in building advanced automation systems.