← Back to QuantegyAI

AI Agents & Tool Use

Eight interactive modules · about 3–4 hours · Course 5 (Inside Large Language Models) is the recommended prerequisite. No coding required.

Track your progress. Sign in to save module completion and your mastery scores across devices. Your progress also saves on this device automatically. Open the portal →

You have chatted with an AI. An agent is the next step: an AI that does not just answer, but acts — it decides to look something up, run a calculation, check a calendar, then uses what it learns to decide its next move. ChatGPT with tools, Claude with computer use, the coding assistants that edit your files — they are all the same idea. A model sits inside a loop, and the loop lets it reach out into the world. This course builds that loop from the bottom up, and every piece runs live in your browser.

This course is extensively hands-on. You will execute an agent's perceive–decide–act loop one step at a time; observe a model invoke a calculator rather than generating the answer directly; interleave reasoning and action across multiple steps with the ReAct pattern; fill a context window and observe earlier information being evicted; provide the agent with long-term memory; have it plan before acting; build a router that selects the appropriate tool for a request; and add the guardrails — step limits, loop detection, human confirmation — that ensure a production agent behaves safely. Each module also presents the corresponding LangChain / OpenAI / Claude tool-use construct for later recognition. Each module concludes with a short mastery check; passing it marks the module complete.

About the agent in this course. Every agent you drive here runs entirely in your browser — no API key, no account, nothing to install, and it works offline. To make that possible, the agent's "brain" (the part that decides what to do next) is a small set of rules standing in for the language-model call a real agent would make. The mechanics you will learn — the loop, tool calls, the ReAct trace, memory, planning, routing, guardrails — are exactly the same as in production systems like LangChain or OpenAI and Claude tool use. We mention this once, here; from now on the modules just focus on how agents work.

The core loop

Module 1

The Agent Loop

What distinguishes an agent from a chatbot? It operates in a loop: perceive, decide, act, observe — repeated. Activity: execute a number-guessing agent one step at a time and observe it converge by reacting to each result. AI anchor: this loop is the structural basis of every agent.

 Module 2

Calling Tools

A model that can only generate text is limited; a model that can act requires tools. Activity: pose a mathematical question to an agent and observe it invoke a calculator rather than generating an answer, then incorporate the result. AI anchor: "tool use" / "function calling" is the mechanism by which ChatGPT browses the web, executes code, and retrieves weather data.

Reasoning across steps

Module 3

The ReAct Pattern

Complex questions require more than one tool call. Activity: pose "how many more people live in Tokyo than in Paris?" and observe the agent interleave Thought → Action → Observation — retrieving each city's population, then computing the difference — until it reaches the answer. AI anchor: ReAct is the pattern underlying nearly every modern agent.

 Module 4

Memory: Context & Recall

An agent can access only the information in its context. Activity: fill a small context window and observe the oldest information being evicted, then provide the agent with a long-term store it can write to and retrieve from. AI anchor: the distinction between the context window and a vector database / memory store.

 Module 5

Planning Before Acting

For a multi-part goal, proceeding one step at a time without a plan is unreliable. Activity: provide the agent with a goal, observe it decompose the goal into a numbered plan, then execute the steps in order. AI anchor: plan-then-execute is how agents handle long, structured tasks.

Making it reliable

Module 6

Routing to the Right Tool

Given a set of available tools, which one is appropriate for the request? Activity: send different questions through a router and observe it dispatch each to the calculator, the lookup, the calendar, or search. AI anchor: tool selection is the decision an agent makes at every step.

 Module 7

Guardrails & Reliability

Agents fail in characteristic ways — non-terminating loops, incorrect tool selection, and unintended actions. Activity: observe an agent enter a failure state, then enable a step limit, loop detection, and human-in-the-loop confirmation and observe it recover. AI anchor: the safety mechanisms every production agent requires.

Project

Module 8 · Project

Build an Agent — and Know When Not To

Assemble every piece — loop, tools, ReAct, memory, planning, routing, guardrails — into one working agent and give it a real task. Then the most important judgment of all: when NOT to use an agent, and reach for a plain function or a single model call instead. A synthesis check ties every module together.

Capstone

Capstone

Build a Concept Manipulative

Put it all together: build a single-page interactive that teaches one AI Agents concept, then submit it for grading and your certificate.
Why this matters This is where everything you learned becomes something that acts. The language model from Course 5 is the brain; this course is the body around it — the loop, the hands (tools), the memory, and the judgment that keeps it safe. After this, agents are no longer mysterious: you know exactly what is happening when an AI decides, on its own, to go and do something.

← Back to QuantegyAI