What is an AI agent?

Definition

An AI agent is a software system in which a large language model (LLM) decides, step by step, what to do next and which tools to use until a task is done. Anthropic describes agents as "systems where LLMs dynamically direct their own processes and tool usage, maintaining control over how they accomplish tasks."

Checked: 2026-10-29 (Beijing time) · Sources

Agent or workflow?

Anthropic groups both under the name "agentic systems" but draws a line between them. Workflows are "systems where LLMs and tools are orchestrated through predefined code paths". Agents choose their own path. Anthropic calls the basic building block an "augmented LLM": a model enhanced with retrieval, tools and memory.

Anthropic's post was published on Dec 19, 2024, and the page itself now says that much of the tooling it describes has changed since then. The workflow-versus-agent distinction is still the clearest short definition we found.

Examples

To build a small agent yourself, start with our tool-calls agent prompt. Before you trust an agent's code changes, see AI code review.

Related tools on this site

Related pages

Sources

We only state facts we checked on the linked sources on the dates shown. Found a mistake? Email jazz.fanpeng@mail.grokbot.com.