1 / 29

Free, step-by-step slide guide

Claude Fundamentalsfrom first chat to real work.

Learn Claude by doing one useful thing at a time. Install the desktop app, choose the right model, set safe defaults, connect live context, and save a workflow you can run again.

10 steps29 slidesAbout 65 minutesNo codingFree account works

The finish line

One workflow you can reuse next week.

Not a feature tour. A completed piece of work, a good example, and instructions Claude can use again.

Progress is saved in this browser.

Free, step-by-step slide guide

Claude Fundamentalsfrom first chat to real work.

Learn Claude by doing one useful thing at a time. Install the desktop app, choose the right model, set safe defaults, connect live context, and save a workflow you can run again.

10 steps29 slidesAbout 65 minutesNo codingFree account works

The finish line

One workflow you can reuse next week.

Not a feature tour. A completed piece of work, a good example, and instructions Claude can use again.

Progress is saved in this browser.

Before you begin

Bring one useful task. Leave with one reusable workflow.

Keep Claude Desktop open beside this guide. You will do each task yourself, then confirm it before continuing.

Bring these two things

• One low-risk file, such as meeting notes or a public report
• One example of a finished result you like

1Claude Desktop installed with a model you chose on purpose
2Privacy and Custom Instructions set for the way you work
3One connector checked against its live source
4A clarified prompt improved with feedback and your own file
5A reviewed Artifact and reusable Project for next week

The guide

Learn Claude by doing one useful thing at a time.

Step 1 of 10 · 5 min

Install the Claude app for Mac or Windows and sign in with your Claude account.

Download Claude Desktop

Claude Desktop keeps chat, files, connectors, and longer workflows in one app. Starting here means every later step happens in the same place.

Do this

1Go to claude.ai/download and choose the version for your computer: Mac or Windows.
2Run the installer, then open Claude Desktop.
3Sign in with the same Claude account you use on the web, or create one if this is your first time.
4Check the plan and tools available to you. The core guide works with a free account, while some models and connectors may require a paid or workspace plan.

Claude Desktop showing the option to open Claude's desktop workspace. — Use Claude Desktop for the rest of this guide so your settings, chats, and connected tools stay together.

You are done when

Claude Desktop is open on your computer and you are signed in.

Step 2 of 10 · 7 min

Find the core chat controls, then learn when to use Sonnet, Opus, Haiku, or Fable.

Learn the four parts of the screen

The chat controls tell Claude where your work belongs. The model picker decides how much speed and reasoning power Claude brings to it. Sonnet is the best starting point for most work; the other models are useful when the task calls for a different tradeoff.

Do this

1Click the large message box. This is where you type and send requests.
2Find New Chat, usually shown as a plus. Use it when you change topics.
3Find your conversation history in the left sidebar. Open any old conversation to continue it.
4Find the plus inside the message box. You will use it later to attach a file or choose a connected tool.

Claude home screen showing the left sidebar, message box, plus button, and model picker. — The labels may move as Claude changes, but New Chat, history, attachments, and model selection remain the core controls.

Step 2 of 10 · 7 min

Find the core chat controls, then learn when to use Sonnet, Opus, Haiku, or Fable.

Use Sonnet as your everyday default

Do this

5Find the current model name near the message box and open the model picker. The exact models shown depend on your plan and workspace settings.
6Choose Sonnet for this guide. It is the best balance of speed and capability for most everyday work.

Remember: Model versions and availability change. Use the models shown in your picker as the source of truth, and start with Sonnet when you are unsure.

Quick model guide

Model	Best for	What to know
Sonnet 5	Everyday writing, analysis, questions, coding, and tool use	Start here. It offers the best balance of speed and intelligence for most work.
Opus 5	Complex research, multi-step reasoning, and demanding professional work	Use it when Sonnet is not getting you far enough. It is more deliberate and uses more of your allowance.
Haiku 4.5	Quick lookups, summaries, formatting, and simple high-volume tasks	It is the fastest option, but it is not the first choice for a difficult or nuanced problem.
Fable 5	Long-running agents, deep reasoning, and your highest-value complex work	It is Claude's most capable general model, but it is slower, uses allowance fastest, and is available only on paid plans.

Step 2 checkpoint

Know where the work goes and which model to use

You are done when

You can start a new chat, find an earlier conversation, attach a file, and explain why Sonnet is the right default for this guide.

Step 3 of 10 · 6 min

Review model improvement and memory, then decide what you will never paste into an AI tool.

Set your privacy boundaries

Claude can only protect information you choose to share. Settings give you control, but your most useful safeguard is a simple rule for sensitive material before real work begins.

Do this

1Open Settings from your name or initials, then open Privacy.
2Review Help improve our AI models and set the switch deliberately. Turn it off if you do not want ordinary chats used for model improvement.
3Find Memory. Depending on your account, it may appear in Settings or Settings → Capabilities. Review what Claude remembers and whether you want it on.
4Use an incognito chat for a temporary conversation you do not want saved to ordinary history or memory.

Claude Privacy settings showing the model improvement control and links to memory preferences. — Claude is rolling out a newer Memory screen, so your labels may differ slightly.

Step 3 checkpoint

Decide what never gets pasted

You are done when

You made a deliberate privacy choice and can state your personal rule for sensitive information.

5Adopt this rule: do not paste passwords, account numbers, private health records, or confidential material you are not authorized to share.

Remember: A useful test: if you would not hand the information to a trusted contractor, do not paste it into a consumer AI account.

Step 4 of 10 · 8 min

Give Claude standing context about who you are, who you serve, and how you want it to respond.

Give Claude the context you should not repeat

Custom Instructions are an onboarding document for Claude. You write them once, and every chat that uses them starts with the role, audience, standards, and reasoning approach you would otherwise have to repeat.

Do this

1Choose where the instructions should apply. Use Profile preferences for broad guidance across Claude, or a Project's instructions when the context belongs to one kind of work.
2Open your chosen instructions area. In a Project, create or open the Project and find its instructions; for broader guidance, open Settings and find Profile preferences or Instructions for Claude.
3Copy the example below and replace the role, audience, tone, and standing rule with your own details.

Claude's Instructions field with the Adaptive Reasoning Protocol pasted in. — Use broad instructions for guidance you always want, and Project instructions for context tied to one workflow.

Step 4 of 10 · Try it now

Add standing context and a reasoning approach

Custom Instructions example

About me: I run a small bakery and write our weekly newsletter.
Audience: local customers, friendly and casual.
How to respond: short paragraphs, no jargon, warm tone.
Always suggest one seasonal item to feature.

Adaptive Reasoning Protocol

Use the Adaptive Reasoning Protocol:

Assess the request type, then apply the appropriate method:

If solving, analyzing, or debugging → Atom of Thought: Decompose into atomic reasoning units. For each atom:

State the logical component
Validate independence
Verify correctness
Then synthesize atoms into final answer.

If explaining, learning, or teaching → Feynman Loop: Explain as if teaching a curious beginner. For each cycle:

Use a concrete analogy
Flag confusion points
Ask questions that reveal gaps
Then compress into a teachable snapshot.

If both are needed → Chain them: First solve via Atom of Thought, then explain the solution via Feynman Loop.

4Add the Adaptive Reasoning Protocol below when you want Claude to choose between problem-solving and plain-language teaching automatically.
5Save the instructions, start a fresh chat that uses them, and ask Claude to summarize what it now knows about how to help you.

Step 4 checkpoint

Claude starts with your working context

You are done when

A new chat reflects your role, audience, response preferences, and reasoning instructions without you repeating them.

Step 5 of 10 · 7 min

Connect one tool you already use so Claude can work from live information you approve.

Connect one live source you already trust

A connector is the bridge between Claude and a tool such as Google Calendar, Gmail, Google Drive, Slack, or Microsoft 365. Once connected, Claude can use the information you authorize instead of relying on copy and paste.

Do this

1In Claude Desktop, open Settings and find Connectors. Your plan or organization may control which connectors appear.
2Choose one low-risk tool you use, such as Google Calendar.
3Select Connect and complete the tool's own sign-in and permission screen. Review the requested access before approving it.
4Start a new chat and ask one question only that connector can answer, such as: "What is on my calendar tomorrow?"

Remember: Start with read-only lookup work. Review Claude's source and answer before you rely on it, and disconnect a tool anytime from its connector settings.

Claude Connectors settings listing available tools and their individual connection status. — Each connector is authorized separately. Connecting Calendar does not automatically grant access to Gmail or Drive.

Step 5 checkpoint

Confirm the connector against its source

You are done when

At least one connector is active, and Claude correctly answered a question using live information from it.

5Check the answer against the source tool. If it is wrong or cannot connect, return to Settings and review that connector's status and permissions.

Step 6 of 10 · Try it now

Start with the problem, not a guessed solution

Do this

1Select New Chat so this task starts with a clean conversation.
2Copy the How Might We prompt below and replace the weekly-update task with something you actually do, or use it unchanged for practice.
3Answer Claude's clarifying questions with real context and constraints.
4Ask Claude to show you the finished reusable prompt, then run it in the same chat with low-risk example or real information.

Remember: Compare the difference: "Write my weekly update" makes Claude guess. "How might we write a weekly update from my meeting notes?" names the input and opens the door to useful questions.

How Might We prompt

How might we write a weekly update from my meeting notes?

Ask me up to three clarifying questions about the audience, format, and what matters most. Then create a reusable prompt I can use each week.

Step 6 checkpoint

Claude asked before it guessed

You are done when

Claude asked clarifying questions, created a reusable prompt, and used it to produce a first answer you can improve.

5Keep this chat open. In Step 7, you will improve the first answer instead of starting over.

Step 7 of 10 · 5 min

Treat the first response as a draft

You do not need to rewrite the original prompt every time. Claude remembers the conversation, so direct feedback is usually the fastest path to a useful result.

Draft

Name what worked

Name what changes

Revise

Save the example

1Name one thing the draft got right.
2Name one thing that needs to change. Be specific about tone, length, structure, or missing information.
3Ask for a revised version in the same chat.

Step 7 of 10 · Try it now

Coach the answer in the same chat

Do this

4When the result is good, tell Claude that this is the version to use as the example next time.

Remember: The result is not the only thing worth saving. A strong example teaches Claude what good looks like on the next run.

Feedback prompt

This is close. Keep the warm opening.

Change the middle so it sounds more personal, remove any generic praise, and make the whole note about 20% shorter.

Give me one revised version only.

Step 8 of 10 · 5 min

Attach meeting notes, a PDF, document, spreadsheet, screenshot, or photo and ask a specific question.

Give Claude the source material

Claude becomes much more useful when it can work from the same source material you use. The safest first task is extracting and organizing information, not making a high-stakes decision for you.

Do this

1Choose a low-risk file such as meeting notes, a public report, or a sample document.
2Click the plus in the message box and choose Add files or photos.

Claude attachment menu with Add files or photos selected from the plus button. — Attach the source and the request together so Claude knows what to do with it.

Step 8 of 10 · Try it now

Ask one specific question about the file

Do this

3Attach the file and send the file-review prompt below in the same message.

File-review prompt

Read the attached meeting notes.

List:
1. Decisions that were made
2. Action items, owners, and deadlines
3. Blockers
4. Unanswered questions

Do not invent missing information. Write "Not stated" when the notes do not say.

Step 8 checkpoint

Check Claude against the source

You are done when

Claude's answer uses the file's actual contents, and you checked it against the source.

4Check at least two details against the source. Correct Claude if anything is missing or unsupported.

Step 9 of 10 · 7 min

Chat is the conversation. The Artifact is the work.

Chat is the conversation. An Artifact is the thing you are producing: a document, checklist, table, or simple tool that stays visible while you request changes.

Chat

The conversation

Questions, feedback, direction, and revision requests.

Artifact

The deliverable

The document, checklist, table, or tool you will actually use.

Claude Artifacts library with a New artifact button and saved creations. — Artifacts keep the deliverable separate from the conversation used to make it.

Step 9 of 10 · Try it now

Turn the findings into a finished deliverable

Do this

1In the chat with your file, ask Claude to turn the result into a polished one-page Artifact.
2If no Artifact appears, open Settings → Capabilities and confirm Code execution and file creation is enabled.
3Ask for one visible change, such as moving decisions to the top or shortening the introduction.
4Use Copy or Download when the Artifact is ready.

Artifact prompt

Turn your findings into a polished one-page Artifact.

Put decisions first, then action items, blockers, and unanswered questions. Make it easy to scan. Do not add information that is not in the source.

Step 9 checkpoint

Make one visible change before downloading

You are done when

You created an Artifact, changed it through the chat, and copied or downloaded the result.

Step 10 of 10 · 8 min

Move from one useful answer to a reusable workflow

One useful answer saves minutes. A reusable workflow saves time every week. Projects keep the instructions, reference files, examples, and related conversations together so you do not restart from zero.

Instructions

How the work should be done

Source material

What the answer must use

Good example

What quality looks like

Fresh input

What changes each week

Step 10 of 10 · 8 min

Put one recurring task, its source material, and your definition of good into a Claude Project.

Create a home for the recurring work

Do this

1Choose one recurring, low-risk task with repeatable inputs and an output you can review. A weekly update is a strong first choice.
2Open Projects in the sidebar and create a project called Weekly Update. Projects are available on free accounts, with a limit on how many you can create.
3Add one example update you like and paste the project instructions below.

Claude Projects gallery with a New project button and existing project cards. — Keep one Project focused on one kind of work so its context stays useful.

Step 10 of 10 · Try it now

Give the Project its operating instructions

Project instructions

Purpose: Turn meeting notes into a clear weekly team update.
Audience: My team. They need the important information quickly.
Output: Decisions, progress, blockers, and next steps in under 300 words.
Quality bar: Direct, specific, and easy to scan. Use my saved example as the style reference.
Safety: Never invent an owner, deadline, or decision. Write "Not stated" when the source does not say.

Your first reusable workflow

How might we turn these meeting notes into a short weekly update for my team?

Ask me up to three questions about the audience, format, and what matters most.

Then create the update as an Artifact. Keep it under 300 words and separate:
- Decisions
- Progress
- Blockers
- Next steps

4Upload fresh meeting notes and run the How Might We prompt.

Step 10 checkpoint

Make the next run better than this one

You are done when

Your Project contains instructions, source material, a finished update, and a good example you can reuse next week.

5Review the result, save the best version as your good example, and update the instructions with anything Claude should repeat next time.

Remember: Use the improvement loop every time: Run → Review → Save the good example → Update the instructions → Repeat.

Claude Fundamentalsfrom first chat to real work.

One workflow you can reuse next week.

Claude Fundamentalsfrom first chat to real work.

One workflow you can reuse next week.

Bring one useful task. Leave with one reusable workflow.

Learn Claude by doing one useful thing at a time.

Download Claude Desktop

Learn the screen and choose a model

Set your privacy boundaries

Add Custom Instructions

Set up connectors

Start a new chat with the How Might We prompt

Improve the answer in the same chat

Work from one of your own files

Turn the result into an Artifact

Save your first reusable workflow

Download Claude Desktop

Learn the four parts of the screen

Use Sonnet as your everyday default

Know where the work goes and which model to use

Set your privacy boundaries

Decide what never gets pasted

Give Claude the context you should not repeat

Add standing context and a reasoning approach

Claude starts with your working context

Connect one live source you already trust

Confirm the connector against its source

Start with the problem, not a guessed solution

Claude asked before it guessed

Treat the first response as a draft

Coach the answer in the same chat

Give Claude the source material

Ask one specific question about the file

Check Claude against the source

Chat is the conversation. The Artifact is the work.

The conversation

The deliverable

Turn the findings into a finished deliverable

Make one visible change before downloading

Move from one useful answer to a reusable workflow

Instructions

Source material

Good example

Fresh input

Create a home for the recurring work

Give the Project its operating instructions

Make the next run better than this one