ModelRefs / Chain of Agents — Agent Pattern
Chain of Agents — Agent Pattern
Agents pass results down a deterministic pipeline; each stage transforms the input for the next. Trades autonomy for reliability and observability.
Overview
A fixed pipeline of specialized agents (e.g., extractor → enricher → writer). Each stage has a narrow prompt and tools and emits typed output for the next. Trades autonomy for reliability and observability.
When to use it: The task has a stable shape and you want reliability over flexibility.
Pattern details
| Pattern class | multi-agent |
|---|---|
| Difficulty | intermediate |
| Autonomy | scripted |
| Also known as | agent pipeline, chained agents |
| Last reviewed | 2026-06-07 |
Known failure modes
- Schema skew — Stage N+1 breaks when stage N changes shape. Mitigation: Version schemas; validate at each boundary.
When not to use it
- Routing dynamic tasks through a fixed chain.
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Chain of Agents — Agent Pattern.
Frequently asked questions
When should I use the Chain of Agents agent pattern?
The task has a stable shape and you want reliability over flexibility.
What are common failure modes of Chain of Agents?
Schema skew
Is Chain of Agents production-ready?
Yes when paired with the safety controls and observability hooks documented on the pattern page.