Glossary Agent
Agent
AgenticAI
Definition
A model running in a loop with tools — it reads, acts, observes the result, and decides the next step — rather than answering a single prompt
The loop is what distinguishes an agent from a chat: it can run a test, read the failure, and change the code without a person relaying each step. What it cannot do is remember past its context window.
Here the agents are Claude Code sessions with the repository, a browser and a shell. They ship through the same pull requests a person would, and the guardrail tests exist because an agent is exactly as careful as what fails. The Dictionary of AI Coding has a longer treatment.