What is AI Agent?

Full Definition

An AI agent is an LLM-powered system that goes beyond single-turn question-answering to autonomously complete multi-step goals. Agents perceive their environment (via tool outputs, web browsing, file access, or API calls), plan a sequence of actions using the LLM's reasoning capabilities, execute those actions, and observe the results in a feedback loop until the goal is achieved. Key components include a planner (often a capable LLM with chain-of-thought), a set of tools (code execution, web search, database queries, browser control), and memory (short-term context window, long-term vector store). Modern agent frameworks such as OpenAI's Operator, Anthropic's Claude Computer Use, and LangGraph enable complex workflows including software development, research synthesis, and business process automation. Managing reliability, cost, and safety in agentic loops remains an active research area.