This is my explanation of what an AI agents is, explained as simply as I possibly can (and as I currently understand it):
An AI Agent is basically a software framework wrapped around (any) LLM, and then empowered to a) be able to make autonomous decisions and b) has access to different tools like;
- the ability to search the web for up to date information
- scan github repositories
- read folder structures
- produce files/ documents
- send e-mails etc.
Here's a metaphore: Your job is to hammer a nail into a plank, but you don't have a hammer or any other kind of tools. An LLM without an agent is sort of in the same position. Sure, it can probably somewhat get the nail in, but the result probably won't be very good.