What is an Agent?

It seems like we’re finally getting a concrete definition of what an “agent” is in the context of AI and large language models (LLMs). From what I’ve gathered, an agent is essentially a system that can autonomously perform tasks by leveraging various tools and APIs orchestrated by an LLM.

I love the recent definition from Simon Willison’s blog post on the topic:

“An LLM agent runs tools in a loop to achieve a goal.”

I thought it fits well with the definition from Hadley Wickham that was shared in a recent Programming with LLMs workshop at Posit Conf 2025:

“Agents are LLMs with a read tool and a write tool.”

They followed it up with some more context:

The “you know it when you see it” definition: autonomous LLMs, long context, minimal intervention

I like the best of both definitions: An agent is an autonomous system that uses an LLM to read from and write using various tools in a loop to achieve a specific goal with minimal human intervention.