AI Agent

A software system that uses a language model to pursue a goal by reasoning, planning and taking actions through tools.

term: AI Agent
category: core
short_def: A software system that uses a language model to pursue a goal by reasoning, planning and taking actions through tools.
long_def: Unlike a single prompt-and-response, an agent runs a loop: it observes, decides on an action (often a tool call), executes it, observes the result, and repeats until the goal is met. Autonomy and tool use are the distinguishing features.
see_also: tool-use agentic-loop mcp
etymology_origin: — verify-against-primary-at-build ↗ https://en.wikipedia.org/wiki/Intelligent_agent — 'agent' in AI predates LLMs (Russell & Norvig, AIMA, 1995); the modern LLM-agent sense has no single coiner
related_to: tool-use agentic-loop mcp agentic-web
contrast_with: Unlike a chatbot, which returns a single response to a prompt, an AI agent runs an autonomous observe-decide-act loop and takes actions through tools to reach a goal.
example: Claude Code and OpenAI's Operator (launched January 2025) are AI agents: they plan, call tools and iterate toward a goal rather than answering once.
source: https://en.wikipedia.org/wiki/Intelligent_agent
status: active
why_it_matters: The AI agent is the first-class visitor the agentic web is built for; understanding it is the precondition for every readiness decision a site owner makes.
sameAs: https://en.wikipedia.org/wiki/Intelligent_agent
bridge_entity: agentic-web
last_verified: 2026-06-15
md_twin: /glossary/ai-agent.md