WS4 · 3 hours · hands-on
Coding Agents - Zero to One
Learn to use a coding agent from scratch. What it is, how to set it up on your laptop, how to give it instructions, how to check its work, what it can do, and where it fails. No coding background needed - the agent writes the code, you direct it. You end the day having driven one through real work - pages, audits, research - on your own laptop.
I’m interestedAnyone who wants to build with AI. No coding background needed.
A laptop and a coding agent account - Claude Pro works, and the pre-workshop email covers the options. Verified at registration.
01 / Learn
Get the
idea.
Open a topic to see what we’ll cover.
01What a coding agent is
An AI that writes and runs code on your machine while you direct it.
02The landscape
Claude Code, Codex, OpenCode, Kimi Code. Same skill, different products. You learn the skill, you pick the tool.
03Your job is approving, not coding
You describe what you want, read what it did, accept or reject.
04Why planning first wins
A wrong plan costs one sentence to fix. A wrong build costs an hour.
05Why fresh sessions matter
A long rambling chat makes the agent worse.
06The standing rules file
A file the agent reads every session, so you state a rule once. Each tool has its own name for it.
07Nothing is ever truly broken
Checkpoints and undo mean any change can be reversed.
08It's not just for code
A coding agent can research, organize files, clean data, draft documents. It is a general worker that happens to live in your terminal.
09What a CLI tool is
A program the agent can run for you: convert files, resize images, pull reports. The agent knows thousands of them. You just ask.
10What MCP is
A plug that connects the agent to outside things: your files, a database, a service. This is how the agent reaches beyond your laptop.
11Where agents fail
What they can't do, and how to spot it before it costs you.
02 / Try it
Put it
to work.
Things you’ll practise during the workshop.
- 01
Set up a coding agent on your own laptop.
- 02
Write a good instruction: what it does, who uses it, what it should look like. Never how to code it.
- 03
Read a change and explain it in plain words before approving.
- 04
Reject a result and re-instruct more precisely.
- 05
Run a task through plan mode and catch a mistake at the plan stage.
- 06
Write your standing rules file.
- 07
Rewind a change you don't want.
- 08
Make the agent explain what it did in plain language.
- 09
Use the agent for non-coding work: research, organizing, summarizing.
- 10
Ask the agent to run a CLI tool you've never heard of.
- 11
Connect one MCP and use it.
- 12
Take a task start to finish alone.
03 / Make it
Your laptop.
Your turn.
Work through these exercises using your own material.
- 01
Build a landing page for your business and see it in the browser.
- 02
Request a change straight, watch it go wrong. Request it through a plan, catch the mistake for one sentence.
- 03
Point it at a codebase you've never seen and have it explain what the code does in plain words.
- 04
Run an SEO audit on your own website.
- 05
Point it at a messy folder and have it organized.
- 06
Give it browser use and have it research something real: competitor prices, a supplier's catalog.
- 07
Have it run a CLI tool you've never heard of: convert files, resize images.
- 08
Solo: pick any task from the day and run it start to finish alone.
Questions, answered
I cannot code. Is this really for me?
Yes. That is the whole design. The famous courses in this space assume coding comfort. This one assumes none. You describe, review, and approve. The agent types.
Which tool do we use?
The skill works the same in Claude Code, Codex, OpenCode, Kimi Code, and the rest. We demo on Claude Code, and everything you learn carries over. You pick your tool.
What will I actually do?
Real work: build a landing page for your business, run an SEO audit on your website, have the agent explain a codebase, organize a messy folder, research competitor prices with browser use. Building a full app comes in the Idea to App workshop.
Do I need the terminal?
No. The desktop apps show every change for approval and preview your work live. The terminal is there when you want it, not required.
What if it breaks something?
Coding agents snapshot your project before changes, and you can rewind to any snapshot. Nothing is ever truly broken.
What comes after?
The habits and the setup go home with you. When you want a real tool on the internet - a database, logins, staff using it on their phones - that is the Idea to App workshop.
Interested in Coding Agents - Zero to One?
Email us about this workshop. We’ll reply with the next dates and details.
Email us about this workshop ↗