🧠 Claude - Full Tutorial for beginners

Note: give it a few seconds for all the videos to load

After the whole Anthropic saga the demand for Claude is insane, so I cancelled all my Sunday plans to build this complete Claude course for ya! Enjoy!

Claude is not just another chatbot. Built by Anthropic — a company whose research is centered on making AI both powerful and honest — it is a tool that consistently rewards people who take the time to understand it. This guide covers all 20 chapters of the Master Claude course: every feature, every workflow, and every insight worth acting on. Whether you are brand new to AI tools or migrating from ChatGPT, this is your complete reference.

What to expect

Chapter 1: The interface and what's worth setting up on day one.
Chapter 2: Which model to use and when.
Chapters 3–4: Web search and Deep Research.
Chapter 5: Custom writing styles.
Chapter 6: Artifacts.
Chapters 7–8: Files and the Excel integration.
Chapter 9: Projects.
Chapters 10–12: Connectors.
Chapters 13–15: Skills.
Chapters 16–17: Claude in Browser.
Chapters 18–20: Claude Cowork.

Chapter 1: Claude Intro

Welcome to Claude — and the first thing worth knowing is that it is not ChatGPT. It is built by Anthropic, a company whose whole research focus is making AI both powerful and genuinely honest. The interface is clean and familiar: sidebar on the left, chat window in the middle, model picker and tools at the bottom. If you have used any AI chat tool before, you will feel at home within about thirty seconds.

Before you dive in, spend two minutes in Settings → Privacy. By default, Claude may use your conversations to improve the product. If you prefer to opt out, a single toggle takes care of it. While you are there, the General tab lets you describe your work and communication preferences — a small investment that pays off every time Claude responds in context rather than cold.

One setting most people skip entirely is Capabilities. This is the control panel for features like Artifacts and Skills — tools we will cover in later chapters. As you progress through this course, this tab is where those features will appear and where you can configure them.

Finally, a note on memory. The free plan does not include Claude's memory feature — each conversation starts fresh. If you are on the paid plan, go to Settings → Capabilities → Import Memory to transfer your context from ChatGPT or Gemini. Claude will ask you to generate a summary in your old tool, then paste it in. From that point on, Claude picks up roughly where your previous AI left off.

Chapter 2: Models Overview

Claude offers more models than most people need, and the confusion is real. Here is the short version that will cover 95% of your use cases.

Haiku is the speed model. Use it for quick, uncomplicated tasks — writing a short email, reformatting a paragraph, answering a simple question. It runs faster and consumes fewer tokens, so if you are hitting daily limits, switching to Haiku for lightweight work stretches your budget significantly.

Sonnet 4.6 is your default for almost everything else. It strikes the right balance between quality and speed. A useful mental model: think of Sonnet as a very capable generalist — someone you trust to handle most of your workload without supervision.

Opus is the specialist you bring in for hard problems. Complex data analysis, multi-step strategy, tasks where the nuance of the reasoning matters as much as the output. The practical advice: start with Sonnet, move to Opus only when Sonnet falls short. You will notice the difference when it matters, and save tokens when it does not.

Extended Thinking is a separate toggle. Leave it off for research and factual lookups. Turn it on when you are brainstorming strategy or working through ambiguous problems — it slows the response but the reasoning quality improves noticeably.

Chapter 3: Web Search + Deep Research — Part 1

Here is something that will earn Claude immediate respect: when it does not know something, it says so. Most AI models will confidently invent an answer. Claude is trained to ask clarifying questions instead — and that behavioral difference makes it dramatically more trustworthy for anything factual.

The fix for Claude's knowledge cutoff is the Web Search tool, accessible from the toolbar below the chat window. Once enabled, Claude searches current sources before responding. The difference between a response with web search off versus on can be significant — especially for anything that has happened in the last year.

Pro tip: if Claude does not pick up the web search tool on its first attempt, start a fresh chat. Once it is working, you will see Claude stepping through multiple sources in real time — you can click in to see exactly what it read and where it found each piece of information.

Deep Research is the paid-tier version of web search. Instead of a single search, it spins up multiple sub-agents — each researching a different angle of your topic simultaneously. Think of it as sending four researchers to different sections of the library at once. It takes 8 to 10 minutes to run, but the output is genuinely comprehensive. Use it for learning plans, competitive research, or any topic where surface-level answers are not enough.

Chapter 4: Web Search + Deep Research — Part 2

When Deep Research finishes, most people scroll straight to the output. Do not. Check the sources first. Every Deep Research session logs the queries it ran and the references it used. A quick scan of those sources tells you whether Claude is reasoning from good information or patching gaps with inference.

The research plan is also worth a look. Deep Research shows you the structure it built: the breadth-first query strategy, the parallel sub-agents it dispatched, and how it synthesized their findings. When something in the output looks off, the plan tells you where the reasoning went sideways.

For lighter research tasks, web search combined with Extended Thinking is often the better choice. You get current information and deeper reasoning, without the 10-minute wait. The prompt matters more here: the more structure you give (role, goal, constraints, output format), the tighter the result.

One useful habit to build: after a Deep Research session generates a long document, prompt Claude to turn it into an interactive artifact (more on this in Chapter 6). A 12-month learning plan is far more usable as a clickable tracker than a wall of bullet points.

Chapter 5: Custom Writing Styles

If you do any regular writing — LinkedIn posts, client emails, internal reports, newsletter content — Custom Writing Styles is the feature that will save you the most time. It is also one of the few things Claude does that no other major AI tool has matched.

The concept is straightforward: you define how you want Claude to write, save it as a named style, and apply it with one click to any future prompt. Instead of explaining your tone preferences every conversation, you do it once and Claude carries it forward consistently.

You have two ways to create a style. Descriptive — write a plain-language explanation of your preferred tone, sentence structure, and audience. Or example-based — paste in a piece of your own writing and let Claude reverse-engineer the style from your sample. The example-based approach tends to produce more accurate results, especially for distinctive voices.

In practice, the difference between styled and unstyled output is immediately obvious. The same prompt run through a custom LinkedIn voice and a default response will feel like two different writers. If you create even one well-defined style and use it consistently, the quality and consistency of your AI-generated writing improves dramatically — with zero extra effort per prompt.

Chapter 6: Claude Artifacts

Most people use Claude in one direction: ask a question, get text, copy it somewhere useful. Artifacts break that pattern. They turn Claude's output into interactive, shareable mini-applications that live inside the interface.

The example that makes the concept click: imagine Claude generates a 12-month AI learning plan. As text, it is useful for about five minutes. As an artifact, that same plan becomes a visual progress tracker with checkboxes, week-by-week structure, and a progress bar. You open it in Claude, check off what you have done, and actually use it over time.

Creating an artifact is as simple as adding the phrase "turn this into an interactive artifact" to any prompt. Claude writes the code, renders a preview in a side panel, and gives you options to publish or share. The more specific you are about what you want, the better — "make it interactive" is a starting point, but "turn this into a visual tracker with checkboxes, monthly progress view, and a completion percentage" is what gets you something genuinely useful.

Any time you find yourself generating content you will need to reference over time — rather than just read once — think about whether it should be an artifact instead of a chat response.

Chapter 7: Files — Part 1

Uploading files to Claude is simple and follows the same pattern you know from other AI tools: hit the attachment button, select your file, and Claude reads and works with the content. PDFs, images, Excel spreadsheets, Word documents — all supported.

What is less obvious — and significantly more useful — is that Claude can generate files from scratch. Not just text you then copy into a document. Actual downloadable files: formatted Word docs, structured Excel spreadsheets with working formulas, properly organized CSV exports.

A single prompt asking Claude to "create an Excel file with engagement data and a Word doc with a beginner's AI roadmap" will produce both, ready to download, without you touching a template or formatting menu. The Word document comes back with proper headings, body text, and structure. The quality is good enough that you spend five minutes editing rather than an hour starting from scratch.

This capability becomes particularly powerful when you are mid-conversation and realize the output would be more useful as a shareable document. Just add a prompt asking Claude to package what it has generated into a file, and it will.

Chapter 8: Files — Part 2 (Excel in Claude)

AI and Excel have had a notoriously rough relationship. Early integrations were mostly cosmetic. Claude's Excel capabilities are a meaningful step beyond that.

The headline feature for Microsoft 365 users: Claude has a plugin that lives directly inside Excel. Once installed, you interact with your spreadsheets through natural language prompts from within Excel itself — no switching tabs, no copying data out and back in.

Even without the plugin, Claude can build fully functional Excel templates from a single prompt. A request for "a personal finance tracker" produced a six-tab workbook: dashboard, budget, savings goals, expense log, net worth, and a how-to guide explaining how to use it. All formulas wired. No errors.

The practical takeaway: before you pay for an Excel template from a creator, try describing what you need and letting Claude build it. For a wide range of common use cases — budgets, trackers, project plans, reporting templates — the output is ready to use immediately, or needs only minor personalization.

Chapter 9: Claude Projects

Projects solve the single biggest friction point in repeated AI workflows: the constant re-explaining. Every time you start a new chat without a Project, you are rebuilding context from zero. Projects eliminate that.

Think of a Project as a dedicated Claude instance that starts every conversation already knowing what it needs to know. Inside a Project, you load three things: files (brand docs, product briefs, customer notes), custom instructions (the prompt framework you would otherwise type every time), and memory (context Claude accumulates automatically through your conversations).

The best use cases are workflows you repeat with consistent context: copywriting in a specific brand voice, customer call analysis, weekly report drafting, content ideation within a defined niche. Once the Project is set up, you just describe the task — Claude handles the rest with full context already in place.

One important distinction from ChatGPT's Custom GPTs: Projects are private, not shareable. But they are available on the free plan, which is a meaningful advantage. If you have any workflow you do more than once a week, it probably belongs in a Project.

Chapter 10: Connectors — Part 1 (Gamma)

Connectors are where Claude starts talking to the rest of your tool stack. Instead of copying content from one app and pasting it into Claude, Connectors let Claude reach directly into external tools and take action — through natural language, inside a single conversation.

The connector library covers the tools you probably already use: Gamma for presentations, Notion for notes, HubSpot for CRM, Canva for design, Figma for product work, Monday.com for project management, Zapier for automation, and many more. Setup takes about two minutes — go to Settings → Connectors, find the tool, authenticate, and it is live.

The Gamma example is the simplest demonstration: you are in a Claude chat, you write "Turn this into presentation slides using Gamma," and Claude calls the Gamma API in the background, generates the deck, and drops a link into the conversation. You never left Claude.

The prompt controls the output. If you want a specific number of slides, a particular structure, or a certain tone, include those details. The default output is presentable but generic. A well-structured prompt produces something you can actually use without heavy editing.

Chapter 11: Claude Connectors — Part 2

The Gamma example is a good introduction, but this chapter is where connectors show their real potential. Using Apify — a no-code web scraping platform with a Claude connector — you can pull live data from Instagram, TikTok, Google Maps, LinkedIn, or dozens of other sources directly into your Claude workflow.

The example: one prompt asked Claude to extract the last 20 Instagram posts from a specific account, calculate engagement metrics, flag the top performers, and generate an Excel report. Claude found the right Apify scraper, ran it, compiled the data — dates, captions, likes, comments, engagement scores — and highlighted the best-performing content. All from a single prompt, with no manual steps.

This is the practical shape of how software interaction is changing. You do not log into Instagram Analytics, export a CSV, open Excel, and spend twenty minutes formatting. You describe the outcome you want, and Claude orchestrates the tools to get there.

Under the hood, this all runs through MCP (Model Context Protocol) — an open-source standard governing how AI tools communicate with external services. The connector ecosystem is built on this shared protocol, which makes it more likely to grow reliably over time.

Chapter 12: How to Connect Connectors

Connecting a tool to Claude takes about two minutes and follows the same OAuth flow you have done for every other app integration. Go to Settings → Connectors, find the tool you want, click it, and log in when prompted. Grant the permissions Claude requests, and the connector appears in your active list.

Once connected, Claude will often use the right connector automatically based on context. If Granola is connected and you ask "what did I cover with David in our last meeting?", Claude reaches out to your meeting notes without being told to. If you want to be explicit, just include the tool name in your prompt.

Managing connectors is worth a moment. Connectors you are not actively using can be disabled without being removed — a toggle in the connector settings. Keeping only the connectors you need active is a good privacy habit: fewer active connections means fewer permissions in play.

One thing to know: some connectors work best through the Claude desktop app, not the web interface. If a connector behaves unexpectedly in the browser, try it in the desktop app first.

Chapter 13: Claude Skills — Part 1

If connectors bring in external data, Skills bring in pre-packaged behaviors. A Skill is a small bundle that combines a detailed prompt with pre-written code, packaged together so Claude can execute a consistent, repeatable task the same way every time — without you having to explain the process again.

The clearest example: the Front-End Slides skill includes both the instructions ("build an animated HTML presentation") and the scaffolding code to produce it. When you type "generate slides for intro to AI for beginners," Claude detects the request matches the skill, loads it automatically, and executes the task using the code already written.

Three benefits stack up fast. Consistency: the output follows the same structure every time. Speed: the code does not need to be rewritten from scratch. Quality: the skill reflects iterated, refined instructions rather than whatever you happen to type in a given prompt. Think of it as handing a new hire a proven playbook instead of asking them to figure things out on the fly.

Skills are distinct from Projects. A Project gives Claude a knowledge base and a persistent persona. A Skill gives Claude a repeatable workflow and code to execute it. If the task is "write marketing copy for my brand," that is a Project. If the task is "generate a formatted Word doc from this transcript every time," that is a Skill.

Chapter 14: Claude Skills — Part 2

Building a Skill sounds more technical than it is. Anthropic has a built-in Skill Creator — a skill that creates other skills — that walks you through the process by asking questions. What should the skill do? What format should the output be in? Who is the audience? After a short conversation, it writes the skill file and packages it as a zip you can download and install.

For people who prefer to browse before building, there is a growing library of community skills on GitHub. Skills for data analysis, brand guidelines, content research, coding workflows, and more. Browse before you build — there is a good chance someone has already created something close to what you need.

One important caveat about community skills: they contain executable code. While the vast majority are benign, it is worth understanding what you are installing before you run it. Anthropic's pre-built skills — available from Settings → Capabilities → Skills — are a safer starting point and cover a solid range of common use cases.

For adding a downloaded skill: go to Settings → Capabilities → Skills, click "Upload a Skill," and select the zip file. Once uploaded, Claude detects when incoming prompts match that skill's purpose and uses it automatically. No further configuration required.

Chapter 15: Claude Skills — Part 3

This chapter is where Skills stop being theoretical and become practical. The course newsletter skill built in the previous chapter gets uploaded, a chapter transcript gets dropped in, and without any explicit instruction to "use the skill," Claude detects the match and activates it automatically.

The output from the skill is structured and consistent: a chapter overview, a video embed placeholder, the newsletter draft, and a prompt asking for tone feedback before continuing. This is the iterative skill loop in action — the skill handles the structural heavy lifting, you handle the directional decisions.

The deeper point is about workflow design. A skill is not just a shortcut — it is the foundation of a repeatable production system. Once you have a skill for newsletter writing, every course you build from here will produce the same quality of output in the same format at the same speed.

The ROI on building a well-designed skill is compounding. The first time you use it, you recover the time you spent building it. The tenth time you use it, you are operating in a different efficiency class entirely. Start small — one skill for your most repetitive output — and expand from there.

Chapter 16: Claude in Browser — Part 1

Claude has a Chrome extension, and it does significantly more than most browser AI extensions. It is not a sidebar chatbot. It is a browser automation agent that can navigate websites, read content, fill forms, extract data, and complete multi-step tasks — while you watch, or while you do something else.

The example from this chapter: finding cheap hotels near SXSW 2026. Claude formulated a plan (look up the dates, navigate to a travel site, sort by price, surface the cheapest options), then executed it in the browser. The user submitted one prompt. Claude did the rest.

This is the practical difference between an assistant and an agent. An assistant gives you information. An agent takes action. With the browser extension, Claude is operating your browser the way you would: identifying elements, clicking, scrolling, reading content, and reasoning about what it finds.

The extension is free to install from the Chrome Web Store and works on the free Claude plan. If you have never tried browser automation before, start with something simple — a product search, a price comparison, pulling data from a site you visit regularly.

Chapter 17: Claude in Browser — Part 2

The hotel search was a demonstration. This chapter gets into something more useful: using the browser extension to analyze your own LinkedIn post performance. Claude navigates to your analytics, reads the data, identifies patterns across your posts, and tells you what is working — in one conversation.

What makes this more than a data retrieval task is the reasoning layer. Claude is not just pulling numbers — it is interpreting them in the context of your question, noticing patterns you might have missed, and giving you a coherent answer about your content performance. Extraction and analysis happen in one step.

Other practical use cases: job searching across LinkedIn, checking competitor pricing on e-commerce sites, reading and summarizing reviews scattered across multiple platforms, and filling out repetitive web forms. If you can do it manually in a browser, there is a reasonable chance Claude can do it for you — especially if it is repetitive.

One note on reliability: browser automation works best on straightforward, mostly-static pages. Complex single-page apps, aggressive bot detection, and multi-factor authentication are genuine obstacles. When it works, it works well. When it does not, Claude will usually tell you why.

Chapter 18: Claude Cowork — Part 1

Cowork is the most powerful feature in this course. It lives in the Claude desktop app (a paid feature), and it combines everything covered in previous chapters — browser automation, file access, skills, connectors — into a multi-step task engine that handles complex, real-world workflows from a single prompt.

The job search example shows what that actually looks like: one prompt asked Claude to search LinkedIn for SDE2 software engineering roles at US companies open to remote hiring in Canada, collect the results, and create a local document with links to apply. Claude built a plan, navigated LinkedIn, extracted the listings, and created the file. The user came back to results.

The invoice extraction example is more representative of daily use: "Go through my Gmail, find all Anthropic invoices, and add the data to my spreadsheet." Claude accessed Gmail through the browser, found every relevant invoice, extracted the amounts and dates, and populated a shared Google Sheet. What would have taken 15 minutes of manual copy-pasting happened with one prompt. This is what agentic AI looks like in practice — not a smarter search engine, but a task executor that handles end-to-end workflows.

Cowork works best when the task is clearly scoped and the goal is specific. "Help me with my email" is too vague. "Go through my Gmail, find all unread emails from the last 7 days with invoice in the subject, and create a table with sender, date, and amount" gives Claude something to execute against.

Chapter 19: Claude Cowork — Part 2

The file renaming example captures the core of what Cowork can do for local work. The task: "Rename all the screenshots in my screenshots folder based on what they contain." Claude looked at each image, understood the content, wrote a descriptive filename, and renamed it — no scripting, no automation tool, just a prompt and a folder.

Scale that up and the value becomes clear. Fifty screenshots, two hundred files with generic names, a Downloads folder that has become a black hole — let Claude do it. The same pattern applies to any repetitive, content-aware file operation: sorting files by category, identifying duplicates, organizing a folder of PDFs.

The website content extraction example pushes further. The task: extract all content from a website, including every linked page, and compile it into a structured document. Claude navigated 30+ pages, extracted the content from each, and assembled it into a clean file automatically — a task that would typically require a custom scraping script.

An honest note on data privacy: Cowork is powerful precisely because it can access and process your actual content. Before feeding sensitive data into any Cowork workflow, check Anthropic's current data processing policy and make an informed decision. The capability is real; using it wisely is your responsibility.

Chapter 20: Claude Cowork — Bonus

This final chapter is the course eating its own tail in the best possible way. The same Cowork capability you have been learning about was used to actually build this course — extracting transcripts from 20 Loom videos, labeling each one by chapter, and compiling them into a document fed into a skill to generate the newsletter you are reading now.

The prompt was straightforward: "Go into my Loom account, find the Claude course videos in chronological order, extract the transcript for each one, label them correctly by chapter, and compile them into a text file." Claude used the browser extension to navigate Loom, worked through each video, and handed back a labeled, organized document.

What this chapter demonstrates is a template for how to think about Cowork as a production tool. You describe the outcome, not the process. You do not tell Claude which buttons to click or which API to call. You explain what you need, and Claude figures out the steps. The gap between "what I need" and "what I have to do to get it" collapses significantly.

There is a certain poetry in the fact that this course was built using the tools it teaches. You do not need a technical background, a development team, or expensive subscriptions to work this way. You need Claude, a clear understanding of what you want, and enough context to let it work. That is the real lesson from all twenty chapters: the clearest constraint on what AI can do for you is not the technology — it is the quality of the task you hand it.

Result

You have now seen Claude from every angle — the models, the tools, the automation, and the agentic workflows that make it genuinely useful in daily work. Pick one feature from this guide, use it on a real task this week, and build from there. Claude rewards consistent use. The more context you give it and the more you invest in setting up Projects and Skills, the more it starts to feel less like a tool you operate and more like a collaborator who already understands your work.

Until next time,
Kushank @PracticalyAI