Worker Agent
A focused agent that does one assigned piece of a larger job and reports back.
Think of it like
A specialist contractor called in for the wiring — does that one part well, then hands it back to the general contractor.
Example
A worker agent is told only “summarize these five sources,” with no view of the wider project, and returns a tidy summary.
How it actually works
Worker agents get a narrow task, a tight prompt, and often no memory of the broader goal — which is a feature, since a smaller scope means fewer ways to wander. Several can run in parallel. Their limitation is exactly their strength: with no big picture, they can optimize their slice in a way that doesn’t serve the whole, so the supervisor’s framing of each task carries a lot of weight.
For product teams
The unglamorous engine of multi-agent work — narrow scope is what keeps them reliable.
For engineers
A scoped agent instance executing a delegated subtask with limited context, returning results to its supervisor.
Related
- Supervisor Agent — The boss that assigns and reviews its work.
- Subtask — The piece of the job it owns.
- Handoff — Handing its result back up.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome