All terms
Agents & tools
AI agent
Also known as: autonomous agent, agentic AI, LLM agent
A system where an LLM autonomously chains reasoning and tool calls in a loop to accomplish a goal — instead of just answering one question.
What it means
An AI agent is a program built around an LLM that decides what to do next, takes an action (calling a tool, writing code, clicking a button), observes the result, and loops until the goal is met or it gives up. The defining trait is autonomy in the loop: the model — not the developer — chooses the next step at each turn.
This is the structural difference from a chatbot. A chatbot answers your message and stops. An agent reads your message, plans, calls a search API, reads the results, calls a calendar API, drafts an email, asks you to confirm, and sends it. The "agentic" part is the model deciding the order and the stopping condition.
The term is overloaded in 2026. Marketing pages call any LLM call with a tool an "agent." Anthropic's "Building effective agents" essay draws a useful line: workflows have predetermined steps with LLMs slotted in, agents let the LLM drive control flow. Most production "agents" are actually workflows with one LLM call wearing a hat. That's usually fine — true agency is harder to debug and more expensive — but knowing which one you're building changes how you design and evaluate it.
Real agentic systems shipping today: Claude Code and Cursor agents (write and run code), Devin (long-horizon coding), OpenAI Operator and Anthropic Computer Use (drive a browser/desktop), Replit Agent (build apps end-to-end). They share a structure: tool-use loop, scratchpad memory, and explicit termination logic.
Example
You ask Claude Code to "fix the failing test in checkout.spec.ts." It reads the file, runs the test, reads the error, edits the code, reruns the test, sees it pass, commits. Five tool calls, no human in the loop after the prompt — that's an agent.
Why it matters
Agents are the 2024-2026 paradigm shift in how AI gets used. Single-shot prompts gave us writing assistants; agentic loops give us workers. Whether they reliably automate 5% or 50% of knowledge work is the open question of the decade — and where most of the AI capex is being aimed.