Building an AI Agent

What an agent actually is under the hood: the model plus tools plus a loop. Function-calling, the reasoning-acting cycle, and where agents go wrong.

1. An Agent Is a Loop The mental model for an AI agent: a language model given tools and a loop. It reasons, decides to call a tool, reads the result, and repeats until done — and you write the loop, the model makes the choices.
2. The Reasoning-Acting Cycle How the loop really runs: function-calling with a JSON schema, the turn-by-turn message exchange between your code and the model, how tool results feed back, and what memory actually means for an agent.
3. Where Agents Go Wrong The honest failure modes: infinite loops, hallucinated tool calls, runaway cost, and unsafe actions — plus the guardrails that contain them: step budgets, validation, approval gates, and observability.