Stage 1 · Lesson 3 of 6
The agent loop
Every runs the same loop. Send the conversation and the list of tools to the model. If the reply contains tool calls, run them, append the results to the conversation and call the model again.
If the reply is plain text, the task is finished. This idea comes from a 2022 research paper called , short for reason and act, where the model alternates between thinking, acting with a tool and observing the result. Modern models do the reasoning internally and return structured tool calls, so you no longer parse text, but the rhythm is identical. You will write this loop by hand in the next stage, in about fifteen lines.
Common trapwriting it without a step limit, so one confused run keeps calling tools until your bill, not your code, stops it.
A question to check yourself, an AI practice session and 2 cheat sheet entries come with the full roadmap. Unlock
Resources for this stage
Start with· follow one of these
- CoursePaidGrokking Modern AI Fundamentals
Design Gurus · Beginner · $72 one-time
Chapter 8, Agentic AI, covers reasoning, planning and memory in agents and ends with a small agent project
- CourseFreeHugging Face AI Agents Course
Hugging Face · Beginner
Free: Unit 1 explains the thought, action, observation cycle and builds a first agent step by step