Back to glossary

Term

AI Agent

An AI agent combines a language model with tools and works toward a goal across multiple steps — typically in a loop of observing, planning and acting.

AI Agent — explained in more detail

An AI agent combines a language model with tools and works toward a goal across multiple steps. At its core runs a loop of observing, planning and acting: the model receives a task, calls tools (file operations, search, shell commands), evaluates the results and plans the next step — until a stop condition is met. The building blocks are therefore an LLM, tool calls, context management and a clear termination criterion.

Example / Practical context

Coding agents like Claude Code, Cursor Composer or Aider follow exactly this pattern: read a file, plan a change, execute the edit, run tests, read the result, fix if necessary. Autonomous browser agents fit the same mold — they navigate pages, click elements, extract data and decide for themselves when the goal has been reached.

A single tool call from an LLM does not make an agent — what matters is the multi-step loop and the model’s own control over the stop condition. Classic workflows with hard-wired steps (such as a “RAG pipeline”) are not agents either, since the model does not determine the order itself.

Entdecke mehr

Themenuebersicht