Agent

Also known as: AI Agent, LLM Agent

An LLM agent uses a language model as a planner/controller that can call tools (search, code, APIs), observe results, and iterate toward a goal — not just answer in one shot.

Agents amplify capability and failure modes: tool misuse, loops, and prompt injection. Keep tools narrowly scoped, add stop conditions, and log every step for debugging.

← Back to AI Glossary