mediumAI Engineering

What is an AI agent and how does the agent loop work?

Q: What is an AI agent and how does the agent loop work?

**TL;DR.** An agent = **model + tools + loop**: the model repeatedly decides an action, the runtime executes it, the observation feeds back — until the goal is met or a **budget** stops it. Everything else in agent frameworks is elaboration on this loop. decide model picks action act runtime runs tool <line class='d-edge'

545 views

Understand the problem

Model + tools + loop: the minimal architecture behind every agent framework.

agentsagent-looptool-usearchitecture

Attempt it yourself

Sketch your approach before reading the solution — that's what interviews test.

Stuck? AI Nudge Available

Get a conceptual hint to guide your logic without spoiling the final implementation.

Study the solution

The solution is waiting

Give it an honest attempt first — then compare your thinking with the full walkthrough.

Read the code

Loop guards around the tool loop

const budget = { steps: 25, usd: 2.0, deadline: Date.now() + 120_000 };
const seen = new Set<string>();

while (true) {
  assertBudget(budget);                          // steps, cost, wall clock
  const action = await decide(messages);
  if (action.type === "final") return action.text;

  const sig = action.name + ":" + hash(action.input);
  if (seen.has(sig)) {
    messages.push(hint("You already tried this exact call. Change approach."));
    continue;                                    // loop detection
  }
  seen.add(sig);
  messages.push(await execute(action));
}

Join the discussion

Discussion (0)

No responses yet. Be the first to share what you think.

What is context engineering and how does it differ from prompt engineering?

What is chain-of-thought prompting and when does it actually help?