The definition of agent is interpreted differently by different people

[joyasree78]

Although there are certain areas people converge, but I see different people interpret agents differently. Should we have a common definition and and a common protocol for Agents. I see Langchain came up with a agent protocol and so as Cisco, which they are calling as Internet of Agents. I would like to hear the community’s perspective on this and also if we should have a common body guiding the definition and the protocol of the agents.

To me , the definition of agent is any software that can generate content or take action based on a configurable level of autonomy through reasoning, planning and leveraging memory of past interactions

[EricGT]

@joyasree78

different people interpret agents differently

First off, the word agent as currently used with LLMs or chatbots is not a new term. The word agent has been around for some time with AI in general.

One of the classic books on AI “Artificial Intelligence : A Modern Approach” by Stuart J. Russell and Peter Norvig (WorldCat) gives

An agent is just something that acts (agent comes from the Latin agere, to do).

As such it is a very open definition.

However there are much more specific examples of agents noted in the book,

02_IntelligentAgents.pdf

A related post with some more information.

Agentic Discovery & Exploration with gpt-4o-mini

First off agents used with AI predate the modern notion of LLM agent. In my view of agents, which predates most of what is now being noted, is to just think of an agent as a subtask that uses AI to accomplish a specific task. In other words if a monolithic prompt does not work, try breaking the process into subtask using a different prompt(s). That is basically it. When I read your question it is like asking in more traditional programming for information on how to create a method or function.…

[aneeshgadhwal]

I would define it as
An agent is an autonomous or semi-autonomous software system that perceives its environment, makes decisions, and takes actions to achieve specific goals. It can interact with other agents or systems, adapt its behavior over time, and perform tasks or generate content with varying levels of independence.

[aneeshgadhwal]

some sources are like to define it more openly as Software_agent.

[mat.eo]

@EricGT

As such it is a very open definition.

Which is the issue, really.

As soon as anything starts to involve monetary value the meaning is pushed, pulled, and distorted into whatever appeals the market.

Once this happens, the resulting definition is one that costs zero brain calories to understand, and hints towards “reduced cost” or “increased profits”.

Which, in this case, AI Agents are a “reduced cost” by replacing humans.

[souly.et.music]

In my opinion semi-autonomous or autonomous programs are not programs but a different kind of entity or digital persons. A program is a program software is software anything that is more autonomous than that, for me it is like a living being, a person.

[Scarletioshub]

A common definition and protocol for agents would help standardize development, improve interoperability, and accelerate adoption across industries. While frameworks like LangChain’s Agent Protocol and Cisco’s Internet of Agents are steps in this direction, having a unified standard from a recognized body (similar to W3C for web standards) could bring clarity. The challenge is balancing flexibility with structure since different use cases require varying levels of autonomy and reasoning. It would be great to see industry collaboration on this—what do others think?