OpenRouter's subagent server tool lets a model hand a self-contained task to another model during a request. The worker receives only the task description supplied by the delegating model and returns its final text. It does not inherit the parent conversation, so the handoff must include the information needed to complete the task.
A developer can select a cheaper worker for routine summarization or formatting while keeping a more capable model in charge of the larger task. That cost distinction requires configuration: without an explicit worker selection, the tool uses the outer request's model.
Terms in this piece · Glossary
multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.
