Claude Code · Beginner
ten modules to get going

How to read this course

Ten modules, fifteen to twenty-five minutes each. Don't try to finish the whole thing in one sitting. One or two modules a day, with your hands on the keyboard, is the pace that sticks. The real checkpoint is module 10 — you hand Claude one task from your actual job, and see what comes back.

No prior knowledge needed. You won't open a terminal. You won't learn git. All you need is the Claude Desktop app and an empty folder you can write to. Pro works, but if you hit rate limits often, Max is worth it.

Module 1 · Open Claude Desktop and see the Code tab

15 min · ★ · No prerequisites

🎯 Goal

Launch Claude Desktop and spot the difference between the Chat tab and the Code tab by eye.

📖 What

• Launch the Claude Desktop app. If you haven't installed it, grab it from claude.ai/download.
• Sign in and check your plan (Pro or Max) from the account menu in the bottom-left.
• Find the Chat tab and the Code tab along the top or sidebar.
• Open the Code tab and point at each region:
  • Parallel Sessions (sidebar, multiple sessions stacked)
  • Prompt input (the large input area at the bottom)
  • Visual Diff pane (where before/after changes show up)
  • Terminal pane (Cmd+` / Ctrl+` to toggle)
  • Preview pane (for HTML, images)
• Feel the split: Chat is "just a conversation." Code is "a conversation that writes files."

💬 Try this prompt

Hello. I'm starting to learn Claude Code today. First, list five things Claude can do in this session, as bullet points.

✅ Checkpoint

Code tab is open and Claude replied in English. You can point at Parallel Sessions, Prompt, Visual Diff, Terminal, and Preview without hesitating.

💡 Tip

Free plan doesn't show the Code tab at all. Pro ($20/month) is the floor. For work use, Max ($100+/month) is what you want. If the tab doesn't appear after signing in, restart the app — that fixes it almost every time.

Module 2 · Make a "Hello World" HTML from an empty folder

20 min · ★ · Module 1

🎯 Goal

Tell Claude what you want in plain English, get an HTML file, open it in a browser.

📖 What

• In Finder (macOS) or File Explorer (Windows), make an empty folder called claude-practice on your Desktop or home directory.
• In Claude Code, hit Open folder and point at that empty folder.
• Send the prompt below and watch hello.html appear.
• Double-click hello.html in Finder to open it in a browser.
• The Visual Diff pane shows a green "new file added" stripe. No need to accept or reject yet — some setups auto-accept.

💬 Try this prompt

Create hello.html with 'Hello, World' shown as an h1. White background, text centered. Keep it simple.

✅ Checkpoint

A browser page shows "Hello, World" in large letters. Finder confirms hello.html sits in the folder.

💡 Tip

Empty folders are the right starting point for beginners. If existing files are around, Claude will read them and sometimes make edits you didn't ask for. Think of it as "a fresh practice folder each time."

Module 3 · Plan mode — see the plan before you commit

20 min · ★ · Module 2

🎯 Goal

Get Claude to describe what it's going to do before touching a single file. Approve, then let it run.

📖 What

• Toggle Plan mode (next to the Prompt input, or Shift+Tab).
• In the same folder from Module 2, send the prompt below.
• Claude returns just a plan — no files yet.
• Read the plan. If it looks right, hit Approve. If not, say something like "make the nav more prominent" and ask for a revised plan.
• Only after Approve does Claude create files.

💬 Try this prompt

Turn this into a 3-page site: Home, About, Contact. Shared nav bar so the pages link together. Simple, calm color palette.

✅ Checkpoint

You saw the plan first, clicked Approve, then files got written. Three HTML pages exist and the nav links work between them.

💡 Tip

Plan mode pays off most on bigger changes. A one-line typo fix doesn't need it. For executive-level work, my default is: Plan mode on, read the plan, approve. Builds the habit.

Module 4 · Visual diff — accept or reject each change

15 min · ★ · Module 3

🎯 Goal

Read a diff before it lands, and pick what to keep.

📖 What

• Target the Home page from Module 3.
• Send the prompt below.
• Visual Diff pane: green = added lines, red = removed lines.
• Read line by line. Reject anything you don't like, accept the rest.
• Reload index.html in the browser and check the change.

💬 Try this prompt

Change the home page h1 to 'My Site'. Switch the background to a warm tone — beige or pale orange.

✅ Checkpoint

You saw the change before it happened. Accept/Reject was your call, and the browser shows the result.

💡 Tip

When you're not sure, reject once and rewrite the prompt. Half-accepting is messier to undo than starting again. Git would help, but you don't need it here — if something goes sideways, delete the folder and redo from Module 2.

Module 5 · Meeting notes to Markdown

25 min · ★★ · Module 4

🎯 Goal

Hand Claude raw meeting notes, get a clean Markdown summary back. This is the single most useful pattern for exec work.

📖 What

• Make a fresh folder claude-practice-memo and Open folder on it.
• Paste real messy meeting notes into the prompt below.
• Open the generated summary.md in the Code tab's Preview or editor.
• Notice the structure: h2 headings, bullets, bold, tables if needed.
• If it's off, iterate: "tighter," "split decisions from action items," and so on.

💬 Try this prompt

Turn the meeting notes below into summary.md.
Three sections: (1) Decisions, (2) Action items (with owner and due date), (3) Open questions for next time.

[paste your meeting notes here]

✅ Checkpoint

The messy notes came out as a readable summary. Decisions are separated from action items, and it's clear who owns what, by when.

💡 Tip

Take this to your next weekly standup. First time it may take thirty minutes. By the third time you'll do it in five. That's the moment Claude stops being a demo and starts being part of your routine.

Module 6 · Watch Claude recover from its own errors

20 min · ★★ · Module 5

🎯 Goal

See the "fail → diagnose → fix" loop happen on its own, so red error text stops feeling like a disaster.

📖 What

• New empty folder claude-practice-python, Open folder.
• First prompt: ask for a simple Python script.
• Second prompt: make it fail on purpose. "Import a library that doesn't exist, and run it."
• Terminal pane shows the error.
• Claude will say something like "that library doesn't exist, let me try another way" and patch itself.
• Keep going until it runs.

💬 Try this prompt

In Python, create list_txt.py that prints every .txt file in the current directory.

(next prompt)
Temporarily add an import for a nonexistent module 'super_nonexistent_lib' and run it. When the error comes up, explain the cause, then fix it.

✅ Checkpoint

You watched the full cycle: error appears, Claude explains, edits the code, reruns, code works. An error from Claude stops feeling like something broke.

💡 Tip

Nine out of ten beginners freeze at the first error and think the tool is broken. It isn't. For Claude an error is just another piece of information. Send one more prompt and it usually sorts itself out.

Module 7 · Getting .gitignore right

15 min · ★★ · Module 6

🎯 Goal

Have Claude write a .gitignore that keeps API keys and other "never commit this" files out, and know why that matters.

📖 What

• Pick any of your practice folders and send the prompt below.
• Claude generates .gitignore. Open it and read what's excluded.
• Ask "why do we exclude .env?" and let Claude connect it to API-key leakage risk.
• Ask how to verify with a command (git check-ignore and similar). No need to run it — the point is knowing Claude will tell you when you ask.

💬 Try this prompt

Create a .gitignore for this folder. Exclude .env, .DS_Store, node_modules, __pycache__.

(then)
Why do we exclude .env files from git? Explain in three lines, mentioning API keys.

✅ Checkpoint

.gitignore exists. You can say in your own words why .env should never land in a repo.

💡 Tip

Key management is the single biggest way executives trip up when they push AI into the team. The API key hygiene page has the full version. For now, "never commit .env" is enough.

Module 8 · Get Claude to write a small Python script

25 min · ★★ · Module 7

🎯 Goal

Run code you couldn't have written yourself, from plain-English instructions. Learn the shape of "ask Claude to process business data."

📖 What

• In a practice folder, first have Claude generate a dummy CSV.
• Then ask for a Python script that reads it and does the math.
• Claude runs it in the Terminal pane and shows the result.
• Iterate: change column names, change what gets summed.

💬 Try this prompt

First, generate 10 rows of dummy data as sales.csv. Columns: date, product, amount. Amount random between 1000 and 50000.

(then)
Write sum_sales.py that reads sales.csv and prints the sum of the amount column. Run it and tell me the result.

✅ Checkpoint

You wrote zero lines of Python, but a CSV got summed and a number printed. "I don't have to be the one typing the code" has landed as a feeling.

💡 Tip

When you try this on real data — sales CSVs, survey results, team lists — watch for PII and confidential fields. Always run on dummy data first to confirm behavior, then put real data through.

Module 9 · Send Claude to the web

25 min · ★★ · Module 8

🎯 Goal

Have Claude fetch information from the outside world. Meet MCP as a concept.

📖 What

• MCP is the standard Claude uses to connect to external data (web, Slack, Drive, and so on). Today you're trying web search.
• Settings → Extensions: check that an MCP Fetch or web search extension is on. If not, ask Claude directly — it'll walk you through enabling one.
• Send the prompt below.
• Claude searches the web and returns a short summary with citations.
• If it fails, paste the error back to Claude and it'll suggest the next move.

💬 Try this prompt

Pick three AI industry news items from the last week. One paragraph each, with source URLs.

(if web search isn't working)
Which MCP server do I need to enable web search? Where in Claude Desktop's Settings is it? Walk me through it.

✅ Checkpoint

Claude pulled information from outside and summarized it. "This tool isn't sealed off from the world" has clicked.

💡 Tip

The intermediate course covers how MCP works. For now, "the connection point for external tools" is all you need to remember. If you want to go deep now, see deep-mcp.

Module 10 · One real task from your job

20 min · ★★★ · Module 9

🎯 Goal

Pull one task from this week's actual work, hand it to Claude, and find one job you'd happily delegate.

📖 What

• Pick one real task from the last seven days:
  • Board or business-plan draft
  • Turning recorded text or rough notes into minutes
  • Email draft (external apology, internal policy update)
  • Spot-checking a budget spreadsheet for anomalies
  • Industry research (competitor news roundup)
• In the Code tab, make a new folder claude-real-work-YYYYMMDD and Open folder.
• Don't fuss over phrasing. Describe the task in plain English and send it.
• Write what happened into result.md: what you asked for, what was good, what wasn't.

💬 Try this prompt (board opening remarks)

Next week I'm opening our board meeting with three minutes of CEO remarks.
Three topics:
(1) This quarter's headline numbers (revenue up 12% YoY, in line with plan)
(2) New business X status (running two weeks behind, root cause is team headcount)
(3) Top priority for next quarter (cost structure review)

Tone: honest about the facts, forward-looking. Draft three versions, each sized to read in about three minutes.

✅ Checkpoint

One real task went through Claude and came back. You've found at least one use case you'd run weekly.

💡 Tip

If the output felt a bit off, you're reading the situation right. Real-world AI work is Claude's 80 points plus your 20 points of editing, not waiting for Claude to hit 100. Don't wait for perfect — layer your judgment on top.

Next Steps

Finishing ten modules already puts you in the top 10% of Claude users among executives at your level. Two paths from here.

1. Intermediate course

course-intermediate.html covers:

• Writing CLAUDE.md so Claude remembers your project's rules.
• MCP in practice: GitHub, Slack, Google Drive, Notion.
• Skills (Agent Skills): reusable specialist tasks.
• Team use: handoffs to assistants and reports, permissions, audit.

2. Pages to keep around

• glossary.html — one-line definitions for Claude, MCP, CLAUDE.md, Skills, Routines, and the rest. First stop when a word loses you.
• deep-code.html — every Code Desktop feature, including Parallel Sessions and Routines that we didn't cover here.
• deep-api-keys.html — the full version of the .gitignore story from Module 7, so nobody on your team leaks a key.