ModelRefs / Multi-Agent Orchestration Stack — Architecture Pattern
Multi-Agent Orchestration Stack — Architecture Pattern
Supervisor routes tasks across specialized agents over a shared event bus with persistent memory. Shared memory, structured handoffs, and bus-level tracing.
Overview
An event-driven topology where a supervisor agent decomposes goals and routes subtasks to specialist agents. Shared memory, structured handoffs, and bus-level tracing.
When to use it: You need cooperative agents to tackle goals that exceed a single agent's competence.
Pattern details
| Pattern class | agentic |
|---|---|
| Difficulty | advanced |
| Topology | event-driven |
| Also known as | agent swarm, supervisor agents |
| Last reviewed | 2026-06-07 |
Known failure modes
- Coordination collapse — Agents repeat work or deadlock. Mitigation: Strict role boundaries + supervisor-level cycle detection.
- Memory poisoning — Bad output contaminates shared memory. Mitigation: Validated writes via critic.
When not to use it
- Free-for-all agent groups without a supervisor or critic.
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Multi-Agent Orchestration Stack — Architecture Pattern.
Frequently asked questions
When should I adopt the Multi-Agent Orchestration Stack?
You need cooperative agents to tackle goals that exceed a single agent's competence.
What are common failure modes of Multi-Agent Orchestration Stack?
Coordination collapse • Memory poisoning
Is Multi-Agent Orchestration Stack production-ready?
Yes when paired with the safety controls and observability hooks documented on the pattern page.