Agent
A software system powered by an LLM that perceives its environment, plans actions, and executes them — usually across multiple steps and tools.
In 2026, "agent" is the shorthand for a piece of software where an LLM is in the driver's seat. It receives a goal, decides what step to take next, calls tools or writes code to take that step, observes the result, and loops until done or stuck.
The line between an agent and a chatbot is whether the system *acts*. A chatbot replies. An agent reads files, browses the web, writes code, sends emails, opens pull requests. The autonomy can be tightly supervised (every step gated by human approval) or fully unsupervised over hour-long sessions.
Most production agents in 2026 sit between those extremes. They plan ahead, run several steps autonomously, and pause for approval before anything irreversible.
Frequently asked
What is the difference between an AI agent and a chatbot?+
A chatbot only replies in text. An agent takes action: it calls tools, reads files, browses the web, or writes code. Agents loop autonomously across multiple steps to complete a goal.
Are AI agents the same as LLMs?+
No. An LLM is the underlying language model. An agent is a system built around the LLM that adds planning, tool use, and a loop that lets it act. The same LLM can power assistants, copilots, and fully autonomous agents.
What does an agent need to be useful in production?+
A clear goal, a set of tools to act with, a way to observe results, and an evaluation loop. Most production agents in 2026 also have memory, error recovery, and human-in-the-loop gates before irreversible actions.