The operator of a wheel or game, usually at a carnival or amusement park. "Office (tip off) the agent the tip is red (crowd is prosperous and eater to spend)."

- american underworld lingo - 1950
In network management, an agent is the server software that runs on a host or router being managed.

Also the dudes in The Matrix with the tie clips and rectangular sunglasses who finally got Morpheus tied to a chair.

Definition of Agents
An agent is usually considered to be an autonomous software program that acts on behalf of its user. But it is impossible to completely define the term "agent", as applied to computer science. The term has been used with a great diversity of meaning, and researchers cannot seem to agree on the definition. What is generally agreed upon, though, is this:

A common classification of agents
Agents can be classified by the following criteria:

More generally, the term agent can be used to mean anything which can act within an environment. In addition to its use in the field of Multi-Agent Systems, the term will often pop up in other fields or languages, such as software engineering and UML. When discussing agents, there are some other terms that often appear as well:

  • World: AKA the environment, means the collection of all objects under consideration, IE, everything that is available to interact with. Often it also includes a set of rules, for example, the laws of physics, that all objects must obey.
  • Object: Anything in the world that can be acted upon.
  • Agent: Anything in the world that can initiate interactions, acts independently, and has motivations.

The important thing here is that the difference between an agent and an object (even though an agent quite often is also an object...) is that an agent acts freely, an object can only act when acted upon.

Suppose Joe, Jane, and a ball are inside a cubical room with no exits. In this case, the agents are Joe and Jane (assuming they are alive and awake). The objects are Joe, Jane, the ball, and the walls of the room. (Joe and Jane are objects in this example, as I will show.) The world is Joe, Jane, the Ball, the walls, the space within the walls, and the laws of physics. These concepts are so basic that we generally don't have words for them outside of specialty fields.

Joe can kick the ball. The ball responds to being kicked, but it'd be most odd if the ball acted on its own. If it did, we would obviously speculate that there was a third agent present (maybe a ghost). Joe can also kick Jane, and Jane will respond according to the laws of physics (respond as an object), as well as responding according to her own motivations (responding as an agent). If, for this example, we ignore the overly cumbersome process of describing sound wave propagation, etc, we will say that the verbal reprimand that Jane gives Joe for kicking her is agent-to-agent communication, which does not involve the world or any object within.

It is feasible, and practical, to consider and study a pure agent system, where the world is empty except for agents who may communicate with each other. You may visualize this as disembodied philosophers chatting endlessly in limbo using telepathy.

Understanding "worlds", "agents", and "objects" by these definitions feels surprisingly natural once they are grokked. This is because, in fact, it is likely the most basic thing that a new mind must grasp, possibly the first steps of mental development that we undergo as newborns.

As was said before, the definitions of these terms vary from person to person and from context to context. Or, different words might be used altogtether. However, if you understand the concepts here you are off to a good start.

In common law, an "agent" is a person who acts as the legal representative of another. A person represented by an agent is called a "principal." The relationship between the two is called an "agency."

In civil law countries (Scotland, France, Germany, Japan etc.), the word "mandate" (or an equivalent in the local language) is used instead.

Who is an agent

This is best illustrated by example:

Agencies are almost always created by consent, although the exact form of consent varies. Sometimes, agencies are formed by a contract, either oral or written; sometimes, an agency must be formed by written contract (for instance, when the agent is empowered to sell the principal's real estate).

Agencies can also be inferred from the conduct of the agent and the principal, even when there is no contract. A related concept is apparent authority, or "agency by estoppel"—if the principal permits the agent to act like an agent.

Another way to form an agency is by "ratification," which is basically consent after the fact—the agent enters an unauthorized transaction, and the principal later agrees to the entire transaction.

The tiny number of non-consensual agencies arise out of necessity. For example, if a child needs medical treatment and their parents are unavailable, the child becomes an agent of the parent for the purpose of obtaining the treatment. Since children cannot enter contracts, the child temporarily becomes an agent of the parents, and the contract for treatment binds the parents, not the child.

Agencies end when the consent (or necessity) ends. Either the agent or the principal can unilaterally terminate the agency at any time. (If there's a formal contract, they may be liable for breach of contract, but that's a different issue.) Agencies will also end when one of the parties dies or goes bankrupt, or when the purpose of the agency becomes impossible, impractical, or illegal (the last scenario often occurs when a license is needed, but denied or revoked).

Duties of an agent

The exact duties of an agent are often determined by the agency contract. If there is no contract, or if the contract doesn't specify duties, then some default duties apply, known as "fiduciary" duties. These can be condensed into duties of care and loyalty.

Care means that the agent must obey the principal's instructions, keep the principal reasonably informed as to what they're doing on the principal's behalf, protect the principal's confidential information, keep the principal's property (including money) separate from their own, and act with the ordinary degree of skill in their field.

Loyalty means that the agent must avoid any conflict of interest, and if they cannot do so, must inform the principal that the conflict of interest exists.

Liability issues

An agent is considered to be an extension of the principal within the scope of their agency. This means that (assuming the principal has consented) they can sign a contract on the principal's behalf, and the principal will be legally bound to enforce the contract. An agent for a corporation will usually sign a contract like this:


By: (signature)
Nate Oostendorp, CEO

An agent for a natural person would sign like this:

Michael Ovitz, agent for Michael Jackson

That's assuming Jackson is a "natural person," of course, but we'll save that discussion for elsewhere.

Besides contracts, a principal can also be held vicariously liable for the torts or crimes of their agent. Criminal liability usually requires the principal's direction, approval, or participation, but can also arise when an employer knows or should know that their employee is committing a crime on the job (for instance, if a low-level manager is violating discrimination laws, the company can be liable).

Tort liability is easier to pin on the principal. A principal can become liable for a tort if they give improper instructions to the agent ("don't bother fixing those brake pads"), if the principal is negligent in choosing the agent ("do you think your membership in al-Qaeda affects your ability to work at Delta?"), or if the principal has a duty to keep tabs on what the agent is doing and fails to do so ("go ahead and transport the safe deposit boxes in your station wagon... we trust you").

Such cases make sense because the agent's negligence is being furthered by the principal's negligence. However, a principal can be liable for their agent's torts even if the principal played no active role in the commission of the tort: this concept is known as respondeat superior, and is worth a node in itself. Suffice it here to say that respondeat superior usually applies in the context of employment: if a commercial truck driver runs somebody over while on the job, the truck driver's employer will have to pay the damages. Respondeat superior comes up in certain other contexts as well.

Note that in many of these cases, the agent and principal can both be held liable under the doctrine of "joint and several liability." This doesn't mean that the injured party gets paid twice: it means that either the agent or the principal may be held individually liable for all the damages, and will then have to dispute apportionment among themselves.


  • Emerson, Business Law (4th ed., Barron's 2004)

A"gent (#), a. [L. agens, agentis, p. pr. of agere to act; akin to Gr. to lead, Icel. aka to drive, Skr. aj. .]

Acting -- opposed to patient, or sustaining, action.

[Archaic] "The body agent."


© Webster 1913.

A"gent, n.


One who exerts power, or has the power to act; an actor.

Heaven made us agents, free to good or ill.


One who acts for, or in the place of, another, by authority from him; one intrusted with the business of another; a substitute; a deputy; a factor.


An active power or cause; that which has the power to produce an effect; as, a physical, chemical, or medicinal agent; as, heat is a powerful agent.

© Webster 1913.

Log in or register to write something here or to contact authors.