AI_SLANG_ENTRY
Subagent Meaning
A specialized AI assistant that a main agent delegates work to, usually so research, testing, review, or another bounded task can happen in a separate context.
What does Subagent mean?
A specialized AI assistant that a main agent delegates work to, usually so research, testing, review, or another bounded task can happen in a separate context.
A subagent is a helper agent with a narrower job. It keeps focused work out of the main conversation, often with its own prompt, tools, context window, and permission boundaries.
Origin and usage
Moved from multi-agent research language into everyday AI builder vocabulary through coding-agent tools and SDKs that let a primary agent spawn or hand off to specialist agents.
Source type: product-term. Last checked: 2026-07-21.
Current coding-agent product term. Anthropic's Claude Code docs provide a clear official reference; OpenAI Agents SDK docs use closely related patterns such as agents as tools and handoffs.
Why builders use subagents
- To keep large searches, logs, or codebase scans out of the main context.
- To give a specialist prompt and tool set to a repeatable task.
- To run review, testing, research, or cleanup as a bounded piece of work.
- To return a compact answer instead of every intermediate file read or command output.
Subagent versus another chat
A subagent is usually part of one orchestrated workflow. The user may not talk to it directly; the main agent delegates work and receives a result.
A separate agent session or teammate is more independent. It may have its own task, user-visible thread, or coordination channel instead of acting as a hidden helper.
Examples
- I sent the log crawl to a subagent so the main thread would not drown in stack traces.
- The reviewer subagent found the risky auth change and returned a short summary.
FAQ
What is a subagent?
A subagent is a specialized AI assistant delegated by a main agent to handle a focused task, often in its own context window with its own instructions and tool access.
Why use subagents?
They reduce context clutter, let teams specialize agent behavior, and make repeated tasks such as research, review, or validation easier to route.
Is a subagent always autonomous?
No. A subagent may work independently on a bounded task, but it is usually scoped by the parent agent's instructions, tools, and permission model.