Multi-Agent Run (sidebar: WORK → Multi-Agent Run) splits one objective across parallel agents, then merges their outputs.
Requires Command plan or above.
Start a run
- Open Multi-Agent Run (
/multi-agent-run) - Click Start multi-agent run
- Describe the objective and select participating agents
- Monitor subtasks until aggregation completes
What to expect
- Sub-agents work on scoped subtasks in parallel
- Results aggregate into a single recommendation — not live shared memory between agents
- Each subtask shows verification consistent with other agent runs
API
Endpoints are under /api/agent-swarm (Command tier). See OpenAPI.