A free, plain-English study guide built from official GitHub and Microsoft docs. No coding background needed. Every term is defined before it is used, one concept at a time, so you actually finish.
Made for people switching into coding and AI, not for people who already have a computer-science degree. Most guides assume you know the jargon and lose you on page one. This one defines every term the first time it appears and moves one concept at a time, so the goal is simple: you finish.
The GH-600, "GitHub Certified: Agentic AI Developer," tests whether you can design, build, and operate AI agents: software that plans a task, uses tools to act on it, and checks its own work.
It is aimed at developers and technical practitioners who want a credential in agentic AI on GitHub, but the underlying ideas are learnable by anyone willing to study, no prior certificate required. This guide assumes you start from zero.
// figures verified against Microsoft Learn before launch
Agents plan, then act, then evaluate. You learn the same way. You move to the next concept only once the last one holds.
A short story gives the concept something concrete to hold onto, before any jargon appears.
The concept stated simply, with every new term defined the first time it appears.
A concrete example you can picture, not an abstraction you have to decode on your own.
A short quiz checks the concept held before the next one builds on top of it.
A real excerpt from Lesson 1.1. Every concept opens with a story, lands the idea in plain words, shows a real example, then checks it, and every fact is traced to its official source.
You message support to return an order. A chatbot replies: "Sure, here's how, open Orders, find the item, click Return..." It hands you the steps.
An agent doesn't hand you steps. It finds your order, makes the return, and emails you the receipt. The chatbot talks. The agent acts. That is the whole idea.
An agent is an AI that takes actions, not just words. You give it a goal; it decides the steps, uses tools to carry them out, and checks the result, on its own. GitHub's own docs say it plainly:
"Copilot cloud agent is an autonomous AI system that can work independently in the background to complete tasks, just like a human developer."docs.github.com/copilot/concepts/about-copilot-coding-agent · fetched 2026-05-27
Tell a GitHub agent: "The login page breaks when a password is too long, fix it." On its own it reads the code, finds the cause, writes the fix, and opens it for your approval.
You gave it a goal, not a checklist. Deciding the steps itself is exactly what makes it an agent.
Answer: B — a chatbot replies; an agent decides the steps and does the work. That is the one line to remember.
Each exam claim links to the official Microsoft Learn or GitHub doc it came from, with the date it was checked.
Terms are defined the first time they appear, before any quiz uses them. No coding or AI background needed.
Updated as objectives and source docs change. You see the same version everyone does, honest gaps included.
No account, no email wall, no paid tier hiding the good parts. Donations optional, nothing gated behind them.
So you always know what to study next. Phases 0 and A are live now. The rest ship as they are built.
This is a free, open project. It gets better when learners tell us what is unclear, what is missing, or what is wrong. Two ways to take part.
Spotted an error, a confusing explanation, or a topic we should add? One email changes the lesson for everyone who comes after you.
hello@agenticlately.comIf the guide helped you, a small tip funds the hours to keep lessons accurate, cited, and current. Pay what you can, or nothing at all.
Support the project