Agent
A model wrapped in a loop that lets it call tools, read the results, and act until a task is done.
An agent is a language model plus a harness that runs the plan → act → observe loop: the model decides, emits a tool call, your code runs it, and the result feeds the next step. Tools, memory, and context management are what turn a model that only talks into one that gets things done.
Worked example: an LLM wrapped in a loop that can call tools — it reads a goal, decides an action (search, run code, call an API), observes the result, and repeats until done, instead of answering in one shot. Gotcha: the model is the easy part; the hard part is the harness — error handling, context management, stopping conditions. Without a step budget or verification an agent loops forever or confidently goes down a wrong path, so ‘agentic’ reliability is an engineering problem, not just model quality.