An artificial intelligence subagent receives a subtask from a parent agent or orchestrator. It may use its own prompt, context, model, tools and execution loop, then return a result to the larger workflow.
Subagents enable specialization and parallel work but also create additional cost and control paths. Governance should attribute their usage to the parent run, limit delegation depth and preserve boundaries around data, tools and side effects.
Acronyms and aliases
AI subagent variantartificial intelligence subagent variantsub-agent variant
Related terms
Frequently asked questions
Why do AI systems use subagents?
Subagents let a system delegate specialized or parallel subtasks while the parent agent coordinates the overall objective.
How should subagent costs be tracked?
Every subagent model and tool call should be attributed to both its own execution and the originating parent run.
