magier Getting started with Claude · 60 min

magier · internal workshop · Wed 17 June

Getting started
with Claude

A 60-minute tour. What it is, the concepts that matter, and what you should build.

Scroll to begin ↓

01 · Mindset

Treat it like a new hire.

It's new, it's powerful, and yes, it can feel intimidating. On day one a new hire gets things wrong. That isn't failure, it's onboarding. You steer, it learns, you go again.

The people who get the most out of it aren't the most technical. They're the ones who keep using it until it feels like second nature.

02 · The landscape

One name, a few different things.

Three ways to use Claude. Same brain underneath, very different jobs.

Claude Chat
The one you already know. A conversation in your browser or app.
  • Drafts, answers, brainstorms, summarizes
  • Reads files you upload, connects to your tools
  • It talks. It doesn't touch files on your computer
Best for thinking and writing
Claude Code
Claude that takes action on your actual machine.
  • Reads and writes real files, builds, runs tasks
  • Chains steps and reaches into your other tools
  • Runs in the terminal, VS Code, or the desktop app
Best for building and automating
Cowork
The same action-taking power, without the techy interface.
  • Same engine as Claude Code, gentler to use
  • Works on files and multi-step tasks for you
  • Built for knowledge work, not just code
Best for getting things done, no terminal

03 · The concepts

The words worth knowing.

The whole vocabulary, in four simple groups.

Where the info lives
Context is what Claude knows now: the brief, the references. CLAUDE.md is your brand guidelines, read every session. Memory carries facts across sessions.
Saved know-how
A skill is a saved recipe. A skill chain is one skill handing the baton to the next. Composed in text, no magic button.
Reaching your tools
MCP / connectors let Claude reach into Figma, Notion, Slack. You control what each one can do.
Doing the work
An agent acts. A subagent is a delegated specialist. Agent teams are a coordinated crew, the advanced end.

04 · How to think about agents

What it needs to do great work.

The one model worth remembering.

1
Goal
What "done" actually looks like.
2
Skills
The know-how for the task.
3
Tools
Access to actually do the work.
4
Context
Your brand, the brief, the project.

Give all four and it runs. Miss one and it flails.

05 · What you can build

Four layers of opportunity.

Each shown with something we've already built here.

I
Information
AI that helps you see: consolidate scattered info.
magier example — placeholder
II
Tool
AI that does one task on demand: translate, audit, reformat.
magier example — placeholder
III
Agent
AI that runs on its own: scheduled, working while you sleep.
magier example — placeholder
IV
Creation
AI that makes the deliverable: sites, creative, content.
magier example — placeholder

06 · Using it responsibly

Use it like a pro.

The biggest risk isn't that it breaks something. It's client work going somewhere nobody can see.

Risky

  • Client data in uncleared tools
  • Connectors hooked to a whole drive
  • Everyone on private accounts
  • Shipping output without a review

Like a pro

  • Approved tools, company accounts
  • Least access a connector needs
  • A human signs off on client work
  • Built your own thing? Tell someone

You stay in control. By default it asks before it acts.

07 · Getting started

Getting it on your machine.

For those who haven't installed it, it's a two-minute setup. Download it, then open it in the terminal, in VS Code, or as the desktop app. Whichever feels least scary.

No need to do it right now. This just removes the last excuse.

08 · Starter kit

Two installs worth grabbing.

One gives Claude discipline. The other gives it taste.

Superpowers
A disciplined workflow: brainstorm, design, plan, build, verify. The "I'm building something" pick.
Frontend Design
Makes Claude produce genuinely good-looking interfaces instead of the generic-AI look.

Inside Claude Code, run /plugin, open Discover, install.

09 · Your turn

What should you build?

Look at your own work through the four layers. Jot one answer to each.

Information
Where is your info scattered across tools?
Tool
What single task do you repeat constantly?
Agent
What do you wish just happened on its own?
Creation
What do you make from scratch that AI could rough-draft?

10 · To take with you

Keep using it until
it's second nature.

You don't need to be technical. You need to manage it well: goal, skills, tools, context.
Install it, add Superpowers and Frontend Design, pick one thing from your list.
Keep going through the rough days. That's the whole game.