ModelRefs / Single-Tool Agent — Agent Pattern
Single-Tool Agent — Agent Pattern
Minimal agent loop with exactly one typed tool; the simplest production-safe starting point. Eliminates routing complexity and is ideal for narrow assistants.
Overview
An LLM with a single, well-described tool runs in a tight call→execute→respond loop. Eliminates routing complexity and is ideal for narrow assistants.
When to use it: You need an autonomous assistant that does exactly one thing well.
Pattern details
| Pattern class | tool-use |
|---|---|
| Difficulty | beginner |
| Autonomy | guided |
| Also known as | one-tool agent |
| Last reviewed | 2026-06-07 |
Known failure modes
- Over-calling — Model calls the tool when prose would suffice. Mitigation: Add 'only call when necessary' to the system prompt.
When not to use it
- Inflating a single-tool agent into a router prematurely.
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Single-Tool Agent — Agent Pattern.
Frequently asked questions
When should I use the Single-Tool Agent agent pattern?
You need an autonomous assistant that does exactly one thing well.
What are common failure modes of Single-Tool Agent?
Over-calling
Is Single-Tool Agent production-ready?
Yes when paired with the safety controls and observability hooks documented on the pattern page.