Skip to content

Pick your style: Denkbaar

Book a conversation

Articles

Knowledge

What is an agent?

An agent is an AI with tools. As soon as it can do something, instead of only answering, you can call it an agent.

A laptop sits between a folder, a magnifying glass, an envelope and a screen with green blocks.
An agent uses tools to do work. Illustration: Denkbaar.

An agent is an AI with tools. A chatbot gives an answer. An agent can also do something: read a file, search, prepare an email, run a test.

Plain technical English An agent is a language model that calls tools. The model picks a tool, reads the result and picks the next step. It stops when it reaches the goal or when it gets stuck.

Why it matters

Everybody says “agent”, and everybody means something else. Simon Willison once collected 211 definitions. If you do not know what an agent is, you cannot choose where to use one, and where not.

A hand rests on a laptop with an email window beside an open folder.
Preparing an email is a task an agent can do. Illustration: Denkbaar.

How it works

Willison’s short definition: an LLM agent runs tools in a loop to achieve a goal.

yes

Goal

Model picks a step

Tool: search, read, write

Result

Goal reached?

Done

The model picks a tool, reads the result, and repeats that until the goal is reached.

Anthropic makes a useful distinction. In a workflow, the steps are fixed in code. In an agent, the model decides the next step by itself. An agent fits open tasks where you do not know the number of steps in advance. Anthropic also warns: an agent costs more, and a mistake in step 2 carries into step 3.

A hand points to a branching diagram beside a straight sequence of steps.
A workflow follows fixed steps; an agent chooses the next step. Illustration: Denkbaar.

What you do

  • Choose a workflow when you know the steps. It is cheaper and more predictable.
  • Choose an agent when the road to the goal is different every time.
  • Give an agent only the tools it needs, and test it first in a sandboxed environment.

What we can do with you: together we decide which part of your work is a fixed workflow and where an agent helps, and we build it. Book a conversation.

Sources

  1. Simon Willison: I think "agent" may finally have a widely enough agreed upon definition to be useful jargon now
  2. Anthropic: Building effective agents

Topics: agentsbasics