Claude Course: Learn Claude AI for Work in 8 Free Lessons

How this Claude AI course works

Take one lesson per sitting, in order, with your own work as the example: a real email, price list or weekly report. That is the point of a practical Claude training course: you finish with work you can use, not notes. You can follow along on claude.ai, in Claude Desktop or in the mobile apps. Every feature, plan and limit below was checked against Anthropic's Help Center and Claude's pricing page on October 25, 2026.

Lesson 1: Chat basics and prompts that work

Goal: get an answer you can use without five rounds of corrections.

Anthropic's prompting guide suggests treating Claude like "a brilliant but new employee who lacks context". A good prompt gives that context up front.

1. Say what it is for. One line on the business, the reader and the result you need.
2. Name the output. Length, tone and structure, such as "five bullets" or "a reply under 120 words".
3. Give the material. Paste the email or attach the file; Claude cannot guess your prices.
4. Ask it to ask. "Ask me up to three questions first" stops it from filling gaps with guesses.
5. Change one thing per reply. "Shorter, and drop the apology" beats a list of six fixes.
6. Save your standing preferences. Rules you repeat, like spelling or tone, go in Settings > Your Claude (Settings > Account in the older layout), where they apply to every conversation.

I run [YOUR BUSINESS] and I need [THE RESULT, FOR EXAMPLE A REPLY TO A CLIENT WHOSE ORDER ARRIVED LATE].
It is for [WHO WILL READ IT] and should make them [WHAT THEY SHOULD DO OR FEEL].
Here is the material: [PASTE THE EMAIL, NOTES OR FIGURES].
Format: [LENGTH, TONE AND STRUCTURE].
Before you write, ask me up to three questions if anything important is missing. Do not invent facts, prices or names.

With a brief like that, the first reply is often a question rather than a draft:

What to check

Why each part of the brief matters is explained in the prompt and context basics of our guide on how to learn AI.

Lesson 2: Projects and files

Goal: give Claude your documents and rules once, instead of pasting them into every chat.

To add a file to one chat, click + in the message box and choose Add files or photos, or drag it in. Claude reads PDF, Word, CSV, text and image files, plus Excel once code execution is on, per Anthropic's file upload article.

A Project is a workspace with its own files, instructions and memory, shared by every chat inside it. Anthropic's Projects article says Projects work on every plan, with up to five on Free.

1. Create it. Open Projects in the sidebar and click + New Project. Claude cannot see the name or description.
2. Add three to five core files. A price list, your service descriptions and one example of work you are proud of.
3. Set project instructions. Click Set project instructions and paste the prompt below.
4. Test it in a new chat inside the project. Ask something only your files can answer. Older chats move in with Add to project.

You are helping [YOUR NAME OR ROLE] at [BUSINESS], which sells [WHAT YOU SELL] to [WHO YOUR CUSTOMERS ARE].
Use the files in this project as the source of truth for prices, policies and product details. If they do not cover something, say so.
Write in [TONE] and [LANGUAGE AND SPELLING]. Keep answers short unless I ask for detail.
Never invent figures, quotes or customer names.

What to check

Claude Code has its own version of project instructions, a file called CLAUDE.md: see what to put in a CLAUDE.md file.

Lesson 3: Skills

Goal: turn a job you explain again and again into a skill that Claude follows on its own.

A skill is a folder with a SKILL.md file: a short description of when to use it, plus the steps. Claude reads only the descriptions until a request matches one. Skills work on every plan, and Anthropic's skills article lists built-in ones for Excel, Word, PowerPoint and PDF files.

1. Turn on code execution. In Settings > Capabilities, switch on Code execution and file creation (called Create and edit files in the newer Settings layout).
2. Switch on example skills under Customize > Skills.
3. Do the job once in a chat, correcting Claude until the result is right.
4. Ask Claude to turn the chat into a skill with the prompt below, then read the draft before you save it.
5. Or upload one: zip the skill folder itself, then click +, Create skill and Upload a skill. Test it with three phrasings you would really use.

Turn the job we just finished into a skill called [SKILL-NAME].
Keep the steps that worked and turn every correction I made into a rule.
Write a description under 200 characters that says when to use it, in the words I would type.
Show me the SKILL.md before you save anything.

Skills you switch on in Claude also load in Claude Code (version 2.1.273 or later) when you sign in with the same account. Here is one at work:

What to check

The full walkthrough, with a SKILL.md you can copy, is in how to create a Claude skill, and our Claude skills directory lists tested ones.

Lesson 4: Connectors

Goal: let Claude read one app you use every day, safely.

Connectors let Claude search, read and act in apps such as Google Drive, Gmail, Slack and Notion. They work on every plan, and Claude reaches only what your own account can, says the connectors article.

1. Open the directory. Go to Customize > Connectors and click + next to Connectors.
2. Connect one app. Read what it can read and write, click Connect and sign in.
3. Switch it on for a chat. Click + in the message box (or type /), hover over Connectors and toggle the app on.
4. Start read-only. Ask for a search or a summary before you ask it to create or send anything.
5. Keep the list short. With 10 or more connectors on, Anthropic suggests setting Tool access to On demand.

On Team and Enterprise plans, an owner adds each connector for the organization first and can block write actions.

claude.com/marketplace/connectors-plugins

Find [THE DOCUMENT OR THREAD] in my [CONNECTED APP] and summarize what changed since [DATE] in five bullets.
List the exact files or messages you used.
Do not create, edit, send or delete anything.

What to check

For how connectors work under the hood, read what an MCP server is, or browse the connectors that work with the Claude app.

Lesson 5: Hand off a whole task (Cowork, now part of Claude)

Goal: describe a multi-step job once and get a finished file back.

On September 16, 2026, Anthropic began merging Cowork, its agent for office work, into Claude itself. There is no mode to pick: you describe the outcome, and bigger jobs come back as documents, spreadsheets with working formulas or decks that open in PowerPoint, per the Help Center article on the change.

It is rolling out to Pro and Max first, and to Team (in beta) and Enterprise where an admin turns it on. If your message box still shows Chat and Cowork, like the one at the top of this page, select Cowork. Tasks need a paid plan.

1. Open Claude on the web, in Claude Desktop or on your phone.
2. Set the permission setting to Manual.
3. Brief it with the outcome, the format and the inputs. Attach files or name the connected app.
4. Approve or deny each action. You can stop or redirect Claude while it works.
5. Open the file and check it. Files Claude creates appear alongside the conversation.
6. Make it repeat. Say how often, such as "every Monday at 9 AM", review the task Claude proposes and click Schedule. Scheduled tasks run in the cloud; one that needs files on your computer runs only with Claude Desktop open.

Use [THE FILES OR CONNECTED APP] to produce [THE OUTCOME, FOR EXAMPLE A ONE-PAGE SUMMARY OF LAST MONTH'S ORDERS].
Deliver it as [A WORD DOCUMENT, A SPREADSHEET OR A SHORT DECK].
Show me your plan before you start, and ask if anything is missing.
Do not send, delete or share anything without asking me first.

What to check

Our guide to Claude Cowork vs Claude Code shows which jobs belong where, and the news story on the merge walks through a weekly scheduled task.

Lesson 6: Claude Code for non-coders

Goal: let Claude work in a folder of your files and save the result there, with you approving each step.

Claude Code is Anthropic's agent for real files in a folder you choose: non-coders use it to turn exports into reports, clean spreadsheets or build a simple page. It comes with Pro, Max, Team and Enterprise, and needs no terminal: in Claude Desktop it is the Code tab.

1. Install Claude Desktop for Mac or Windows (Linux is in beta), sign in and click Code.
2. Choose Local and open a test folder with copies of real files, never the originals.
3. Check the permission mode in the row under the prompt box and pick Manual.
4. Ask what is in the folder, then give it one real job with the prompt below.
5. Read each approval before you answer, and decline when something looks wrong, with a note on what you want instead.

Look at the files in this folder and tell me in plain words what each one is.
Then [THE JOB, FOR EXAMPLE TURN CLIENT-FEEDBACK.CSV INTO A ONE-PAGE MONTHLY REPORT] and save it as [FILE NAME].
Ask before you create, change or delete anything, and tell me how you checked the numbers.

We ran this job on a made-up caterer's feedback export. In Manual mode, Claude Code asked before running its calculation and before saving the report, and its totals matched our own count: 22 events, average rating 4.1.

Prefer the terminal? The Claude Code quickstart installs it with one command on Mac, Linux or WSL (Windows uses irm https://claude.ai/install.ps1 | iex in PowerShell):

$ curl -fsSL https://claude.ai/install.sh | bash
$ cd ~/Desktop/test-folder
$ claude --permission-mode manual
$ 

What to check

Fifteen ready-made jobs are in Claude Code for non-developers, install help is in how to install Claude Code, and a first weekend project is in how to use Claude Code.

Lesson 7: Plans, models and limits

Goal: pick the plan that fits your work and stop running out of usage mid-task.

Prices are from Claude's pricing page and the Help Center's Max plan article on October 25, 2026, before tax.

The current models are Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1. Free gets Sonnet and Haiku, Pro adds Opus, and Fable is available on Pro only through paid usage credits and on Max for up to half of the weekly limit. Anthropic's Claude Code usage guide gives a rule that works in chat too: Sonnet for most jobs, Opus for hard problems (it uses more of your limit), Haiku for quick ones.

claude.com/pricing

Every plan has a usage limit that resets on a rolling five-hour window, and paid plans add weekly limits. There is no fixed message count: long chats, big files, tools, the model and the effort level all change what a message uses. The web, desktop, mobile and Claude Code share one pool, and on paid plans Settings > Usage shows your bars.

1. New topic, new chat. Long conversations cost more with every message.
2. Reused documents go in a Project, where cached content counts less.
3. Batch related questions and switch off connectors a chat does not need.
4. Match model and effort to the job. Routine work rarely needs Opus at high effort.

Summarize this conversation as a brief I can paste into a new chat: the goal, the decisions we made, the facts and figures we settled, and what is still open. Keep it under 200 words.

What to check

For a plan-by-plan breakdown see Claude Code pricing, and for what to do when you hit a wall, Claude Code usage limits.

Lesson 8: Put it together in one weekly routine

Goal: one recurring job that runs on a Project, a skill, a connector and a schedule.

1. Pick one boring, checkable job, such as a weekly sales summary.
2. Run it by hand twice inside the project, and fix the skill after each run.
3. Schedule it with the prompt below, on a paid plan.

Every [DAY] at [TIME], use the [SKILL NAME] skill on [THE SOURCE IN YOUR CONNECTED APP] and save the result as [A DOC OR A SPREADSHEET] in this project.
Put anything unusual at the top. If data is missing, say so instead of guessing.
Never send, share or delete anything.

What to check

What to do next

Run lessons 1 to 4 this week on one real piece of work, then build the weekly routine from lesson 8 over the next two weeks. If lesson 6 is where you want to go further, our Claude Code course, open for pre-order in early access, takes people who have never written code from setup to real projects step by step. Either way, keep the loop this Claude AI training is built on: brief clearly, approve what matters, and check the result yourself.