AI Agent
Also known as · agent · agentic ai · agentic
An AI system that can take actions and use tools to accomplish a goal, not just chat.
An AI agent is a system built around a language model that can do more than respond — it can plan, call tools (search the web, run code, query a database, send an email), observe the results, and decide what to do next, looping until a goal is met. The model becomes the 'brain' driving a feedback loop of actions.
This is the shift from AI that answers to AI that acts. Coding agents that edit and test code, research agents that gather and synthesize sources, and assistants that manage calendars and inboxes are all examples.
Agency raises the stakes on reliability and safety: an agent that can take real actions can also cause real harm if it's hijacked or makes a bad call, which is why tool permissions, human approval for consequential steps, and defenses against prompt injection matter.