Introduction
Since I started coding 30 years ago, the tools have changed many times, but nothing has changed them as quickly or as deeply as AI. The way software gets made in 2026 looks very different to how it looked even three years ago, and the biggest shift is something called agentic coding.
If you are learning to code right now, that is genuinely exciting news. You get to learn the fundamentals of programming and the tools that professionals are rebuilding their whole workflows around, at the same time. This guide explains what agentic coding is, and points you to the other guides in this series so you can get set up and start experimenting.
What is agentic coding?
You have probably used an AI chatbot like ChatGPT or Claude. You type a question, it types an answer. That is useful, but it is a conversation, nothing more. The AI cannot do anything.
An agentic coding tool takes the same kind of AI (called a large language model, or LLM) and gives it hands. It runs on your computer, usually in the terminal, and it can:
- Read your code to understand your project.
- Write and edit files to make changes.
- Run commands to test whether those changes actually work.
- Keep going in a loop: trying something, checking the result, and correcting itself until the job is done.
That loop is what makes it "agentic". Instead of just suggesting code for you to copy and paste, the AI acts as an agent working on your behalf. You describe what you want, and it goes and does the work, showing you everything along the way.
Should a beginner be using one?
Yes, but deliberately. There is a trap here: if you ask an agent to do your learning exercises for you, it will happily do them, and you will learn nothing. That would be like joining a gym and paying someone to lift the weights.
Used well though, an agent is one of the best learning tools that has ever existed. It is a patient, always-available expert that can explain any line of code, answer any "but why?" question, and show you how a professional would approach a problem. Our advice:
- Write your own solutions to exercises. The struggle is where the learning happens.
- Ask the agent to explain, not to solve. "Why doesn't this work?" teaches you more than "fix this".
- Watch it work on throwaway projects. Seeing how it breaks problems down is a lesson in itself.
Where to start
We have written this series so you can work through it in order:
- Key Agentic Coding Concepts explains the ideas you will bump into constantly: models, tokens, context, and why these tools sometimes get things confidently wrong. Start here if words like "model" mean nothing to you yet.
- Installing an Agentic Coding Platform walks you through installing OpenCode, the tool we recommend, and getting it working with a free AI model. You can be up and running without spending anything.
You will also want a code editor installed so you can read and edit code yourself. If you have not sorted that yet, see our guides on using a code editor and installing VS Code.
A quick word on the pace of change
Everything in this space moves fast. Tools rename themselves, free tiers appear and disappear, and the "best" model changes every few months. We keep these guides updated (check the "Last updated" date at the top), but do not be surprised if a screenshot looks slightly different by the time you follow it. The fundamentals, which are what these guides focus on, change much more slowly than the logos do.