ModelRefs / Single Agent Stack — Architecture Pattern
Single Agent Stack — Architecture Pattern
One reasoning agent with a registered tool set, short-term memory, and tracing. The canonical entry point for agentic systems.
Overview
A single LLM-driven agent calling vetted tools in a ReAct-style loop with short-term memory and per-step tracing. The canonical entry point for agentic systems.
When to use it: You need autonomous tool use behind one decision-maker.
Pattern details
| Pattern class | agentic |
|---|---|
| Difficulty | intermediate |
| Topology | monolith |
| Also known as | solo agent, react agent stack |
| Last reviewed | 2026-06-07 |
Known failure modes
- Infinite loop — Agent never decides to stop. Mitigation: Hard step cap + budget.
- Tool misuse — Wrong arguments or wrong tool. Mitigation: Strict schemas + retries + tool descriptions.
When not to use it
- Letting the agent register new tools at runtime.
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Single Agent Stack — Architecture Pattern.
Frequently asked questions
When should I adopt the Single Agent Stack?
You need autonomous tool use behind one decision-maker.
What are common failure modes of Single Agent Stack?
Infinite loop • Tool misuse
Is Single Agent Stack production-ready?
Yes when paired with the safety controls and observability hooks documented on the pattern page.