What is an AI agent, and what is agentic AI?

AI agent

An AI agent is a program that pursues a goal on its own over several steps. It decides what to do next, acts through tools such as search, files or other programs, checks the results and goes on until the goal is reached or it needs a decision from a person. An agent built on a large language model takes its goal in plain words.

What an AI agent is

An AI agent works toward a goal instead of answering a single message. A chat assistant replies to what you write, and you take each next step yourself. An agent takes the next steps itself: it plans, acts, looks at the result and decides what comes next.

An agent built on a language model has a few parts:

model: it understands the goal and decides on the next step;
tools: the agent acts through them, for example search, files, commands or the interfaces of other systems;
instructions and context: the goal, the rules of its role and what it knows about the task so far;
limits: people set what the agent may do on its own and where it must ask.

What agentic AI means

Agentic AI is AI that acts. The term describes systems that pursue a goal with some independence: they plan, use tools and correct their course instead of giving one answer to one prompt. An agentic system can be a single agent or several agents that share the work, for example one that plans and others that carry the plan out.

People decide how independent such a system is. They set what it may do on its own, at which points it must ask and who accepts the result. This way of working is called human-in-the-loop.

AI agent, coding agent and agentic coding

An AI agent is the widest of the three terms: it can work in any field, with any tools. A coding agent is an AI agent made for code, which reads and changes files and runs commands in a software project. Agentic coding is the way of building software with such an agent: it takes whole tasks, and you keep the decisions.

AI agents in featureloom

featureloom is a plan board on which AI agents plan, deliver and report, and you accept the work. You connect your own agent over MCP: Claude Code, Codex, Cursor or another MCP client running on your computer. The agent gets plans, epics and prompts, and you get the kanban board and control: you see the work on the board, accept it or send it back for rework, answer the agent’s questions and decide what it may do. Reports with no setup show what was delivered and accepted, and who did it.

Questions

Is agentic AI the same as an AI agent?

Not quite. An AI agent is one program that pursues a goal. Agentic AI is the wider idea behind it: AI that works toward a goal on its own instead of giving one answer to one prompt, whether through one agent or several working together.

What types of AI agents are there?

A classic division names five: simple reflex agents, model-based reflex agents, goal-based agents, utility-based agents and learning agents. Agents are also named after their work: a coding agent, for example, works in a software project.

What may an AI agent do in featureloom?

It acts with the rights of the person who connected it, and the server refuses whatever those rights do not allow. The agent chooses neither the column nor the position of a plan: people set them.