← Back to Glossary
Agent
Phase 04 · Agentscore-concept

An LLM-driven system that reasons, plans, and takes actions (via tools) to achieve a goal autonomously.

Why it exists

Standard LLM calls are stateless prompt-in answer-out. Agents put the model in a feedback loop with tools and memory to solve multi-step goals.

How it works

Executes loop: Observe state -> Think (ReAct) -> Call tool -> Observe output -> Repeat until termination condition is met.