In the first module, we cover the fundamentals: how an agent differs from a GPT chat, how the perceive→think→act cycle is structured, and what minimal code is needed to get started. By the end of the module, you'll have a working "hello-world" agent that can call tools.
Why a ChatGPT web interface with an attached API is still not an agent. What needs to be added for an LLM to become an agent. Using the "buy coffee" example — where a chatbot gives up, but an agent succeeds.
The main pattern of agent operation. How an agent decides what to do at each step. Tool calling in Claude/OpenAI API. What an "agentic loop" is and how to stop it correctly.
Environment setup (Python, Anthropic SDK). The first working agent: can read files and count. We run it locally, view logs, and analyze what happens in each iteration.