The ultimate beginner’s guide to Claude AI

Claude AI has grown from a simple chatbot into a full productivity and building platform. You can use it for quick answers, deep research, document work, automation, and even to build real tools that run on your computer. This guide walks you through Claude step by step, from your very first prompt to advanced workflows with projects, connectors, Cowork, and Claude Code.

What Claude AI is (and why people like it)

Claude is an AI assistant created by Anthropic. If you’ve used tools like ChatGPT, the basic idea will feel familiar: you type in natural language, and Claude replies, writes, analyzes, and builds things for you.

Claude has a strong reputation for three reasons:

• It writes in a very natural, human-like style.
• It’s excellent at working with your documents, spreadsheets, and images.
• It’s designed for real work: research, planning, analysis, and building artefacts, not just quick one-off answers.

Claude is available in a free plan, with paid tiers (like Pro) that unlock higher usage limits and advanced features such as deep research and autonomous work modes.

Getting started with Claude

To start using Claude, go to claude.ai in your browser and create a free account using Google or email. You’ll be asked for basic details like your name, birthday, and what kind of work you do.

During setup, you can choose whether your chats are used to improve Anthropic’s models. You can opt in or out based on your privacy preferences. Once you’re through onboarding, you land in the main Claude interface: a simple chat window where you can start typing.

Claude is also available as a mobile app (iOS and Android) and a desktop app (Mac and Windows). All your chats, projects, and settings sync across devices.

Level 1: Talk to Claude with better prompts

Everything in Claude starts with a prompt—the message you send it. The difference between a weak prompt and a strong one is the difference between a generic reply and something that feels tailored and useful.

The three-part prompt: context, task, format

A powerful way to structure your prompts is to always include three parts:

1. Context – Who you are, what you’re working on, and any relevant background.
2. Task – Exactly what you want Claude to do.
3. Format – How you want the answer: tone, length, structure, bullet points, table, etc.

For example, instead of “Write a product description for a candle,” a stronger prompt might be:

“I run a small handmade soy candle shop. Our customers care about cosy homes and natural ingredients. Write a warm, personal product description under 100 words for our new ‘Autumn Cabin’ candle. Include one short sentence that makes you want to light it tonight.”

Same AI, completely different result.

Reply and refine instead of starting over

Claude remembers everything in the current conversation. Treat its first answer as a draft, then reply with adjustments:

• “Make it half as long.”
• “Keep the mention that it’s hand-poured.”
• “Change the tone to more professional.”

This back-and-forth is where most of the quality comes from. Don’t rewrite your whole prompt every time—just refine.

Choosing the right model and effort level

Claude offers different models (for example, Haiku, Sonnet, Opus, Fable) with different speeds and capabilities. The naming and options may change over time, but the pattern stays the same:

• Use the default model (often Sonnet) for everyday tasks.
• Switch to a more powerful model (like Opus or Fable) for complex, high-stakes work: long contracts, detailed analysis, important writing, or multi-step planning.
• Use a lighter model (like Haiku) when you need fast answers to many small questions.

There’s also an “effort” or “thinking” setting. Higher effort makes Claude think longer before answering, which is useful for multi-step reasoning or when accuracy really matters.

More powerful models and higher effort consume your usage allowance faster. Start on the free plan, and only upgrade when you consistently hit limits or need advanced features.

Level 2: Give Claude your world (files, images, web)

Once you’re comfortable chatting, the next step is to let Claude work with your actual information: documents, spreadsheets, PDFs, and images, plus the live web.

Use separate chats for separate topics

As a simple habit, use one chat per topic or project:

• Same task, same chat.
• New topic, new chat.

This keeps context focused and reduces usage compared to one endless conversation. Claude will automatically name each chat based on your first message so you can find it later in the sidebar.

Uploading and working with files

You can attach files directly in a chat: PDFs, Word docs, spreadsheets, and more. Claude will read them in seconds and can then summarize, analyze, or transform them.

Examples of what you can do:

• Summarize a long contract or bank statement in plain language.
• Extract key numbers from a supplier price list and calculate the cheapest option per unit.
• Turn a research paper into a bullet-point study guide.
• Convert messy data into a clean table.

A powerful way to understand this feature is to upload the most boring or confusing document in your life and ask Claude to explain it.

Understanding and improving images

Claude can also see and reason about images. You can upload product photos, screenshots, or even handwritten notes. For example, you might ask:

• “What would you improve about this product photo to make it sell better online?”
• “Read these handwritten notes and turn them into a clean outline.”

Web search vs deep research

Claude can access the live web when web search is enabled. This is useful for:

• Checking current trends in your industry.
• Getting up-to-date information on tools, prices, or regulations.
• Finding and citing sources.

When Claude searches the web, it shows you where the information came from. Make it a habit to check those sources whenever facts matter.

On paid plans, you can also use a deeper “research” mode. Instead of a quick search, Claude reads many sources and produces a structured, cited report—almost like having a research assistant. This is ideal for big decisions (career moves, large purchases, starting a business) or serious research projects.

Level 3: Make real things (files and interactive artefacts)

Up to this point, everything Claude produced lived inside the chat. At level 3, Claude starts creating real, usable outputs: downloadable files and interactive artefacts.

Generating documents, spreadsheets, and presentations

Claude can generate:

• Excel spreadsheets with working formulas (budgets, inventory trackers, planners).
• Word-style documents (reports, resumes, lesson plans, manuals).
• Slide decks and presentations.

For example, you might ask Claude to:

“Create an Excel spreadsheet to track my inventory. Include columns for product name, cost to make, selling price, profit margin (calculated), and stock. Fill in realistic sample data.”

Claude will build the file, which you can then download and open in Excel or another office suite.

Artefacts: live pages and mini tools

When you ask Claude to build something visual or interactive, it can open a side panel and construct it live. These outputs are called artefacts.

Examples include:

• Landing pages and simple websites.
• Quizzes and forms that respond to your choices.
• Small interactive tools, dashboards, or widgets.

You can then refine them in conversation: change colors, layout, copy, or behavior. It’s like having a designer and front-end developer working beside you.

If you’re interested in going deeper on spreadsheets specifically, there’s a dedicated guide on how to use Claude for Excel that walks through more advanced data and formula workflows.

Level 4: Organize work with projects and memory

As you create more chats and artefacts, your work can start to feel scattered. Level 4 solves this by giving Claude structured memory: projects, instructions, and personal memory.

Projects: shared context for a topic

A project is a workspace where related chats live together and share the same background knowledge. Within a project you can:

• Add files to “knowledge” so every chat in that project can use them without re-uploading.
• Set project instructions that describe the situation, brand, tone, and goals.

Once set up, you can start a new chat in that project with a very short prompt, and Claude will automatically draw on the project’s files and instructions. This is a big step up from one-off chats and is ideal for ongoing work like a business, course, or research topic.

Global instructions and personal memory

Beyond projects, Claude can remember things about you more generally.

Global instructions let you tell Claude how to behave in all chats:

• How to address you.
• Your work context (e.g., “I’m a marketing manager in SaaS”).
• Preferences for tone, level of detail, and formatting.

Memory is what Claude learns from your conversations over time. It may remember:

• Your ongoing projects and brands.
• Your tools and workflows.
• Personal preferences you mention repeatedly.

You stay in control. You can view, edit, or delete memories, and you can turn memory off entirely. There’s also an incognito chat mode where nothing is saved or used for training.

Importing memory from other AI tools

If you’ve used another AI assistant (like ChatGPT) and it has a memory or personalization feature, Claude can import that information. You’ll be guided through copying a summary of what the other assistant knows about you and pasting it into Claude’s memory. This feature is experimental, so expect to review and edit the imported details.

You can also enable an option that lets Claude search across your past chats for relevant context, giving you a more “joined-up” experience across conversations.

Using Claude on mobile and desktop

The mobile app is ideal for capturing ideas, snapping photos, and asking quick questions on the go. You might, for example, photograph a product and ask for a suggested price, or dictate an idea while walking.

The desktop app puts Claude a shortcut away all day and is especially important for advanced features like Cowork and Claude Code, which need local access to files and tools.

Level 5: Connect Claude to your apps with connectors and skills

Until now, Claude has mostly talked about your world. At level 5, it starts working inside your existing tools using connectors and skills.

Connectors: plug Claude into your calendar, files, and more

Connectors let Claude securely access other apps you use, such as:

• Google Calendar
• Gmail
• Google Drive
• Slack
• E-commerce and marketing tools

You connect an app once, grant specific permissions, and then you can simply ask Claude to do things like:

• “Look at my calendar for next week and tell me which afternoon has the most free time for deep work.”
• “Search my Google Drive for all documents about [project] and summarize the key decisions.”

You can disconnect connectors at any time, and Claude will ask for permission before performing sensitive actions.

Real example: drafting campaigns in your email platform

Once connected to an email marketing platform, Claude can go beyond writing copy in chat. It can log into your account via the connector, create a new campaign draft, set the subject line and preview text, and paste in the full email content—ready for you to review and send.

Over time, it can also analyze your campaign history, identify what’s working, and suggest improvements.

Skills: reusable “recipes” for specific jobs

Skills are like pre-written instruction sets for Claude. Each skill is essentially a text file that tells Claude exactly how to perform a certain type of task, including steps, formatting, and quality standards.

You can browse a library of skills and install the ones you need—for example, a skill that creates beautifully themed slide decks or polished reports. Once installed, you can:

• Trigger a skill by typing “/” and selecting it from a list.
• Let Claude automatically use it when your request matches that skill.

Skills are especially powerful when combined with artefacts and projects, giving you consistent, high-quality outputs without rewriting instructions every time.

Level 6: Let Claude work while you’re away (Cowork and scheduling)

At level 6, Claude stops being something you talk to constantly and starts becoming something that quietly does work for you in the background.

Cowork: autonomous work on your files

Cowork is a mode in the desktop app where you give Claude a task and a folder, and it goes off to do multi-step work on your computer. You can choose how much freedom it has:

• Manual approval – Claude asks before every action.
• Automatic approval – Claude acts on its own but still runs safety checks.
• Skip approvals – Claude never pauses to ask (best only once you fully trust the workflow).

Before starting, you choose which folders Claude can access. It can only read and write files in those locations; the rest of your system stays off-limits.

Typical Cowork tasks include:

• Reading a folder of receipts, extracting dates and amounts, and building a clean expense spreadsheet.
• Renaming files based on their contents.
• Cleaning and organizing documents into sensible structures.

While Cowork runs, you can see a progress list of the steps it’s taking. For your first runs, it’s wise to stay on manual approval so you can see exactly what it’s doing.

Scheduled tasks: recurring work without supervision

Scheduled tasks let you describe a recurring job once and have Claude run it on a timetable. Examples:

• A weekly industry trend report delivered every Monday morning.
• A monthly summary of your calendar and tasks for planning.
• Regular checks on a budget or KPI spreadsheet.

You can specify:

• What the task should do.
• How often it should run (daily, weekly, monthly).
• What time of day it should run.
• How detailed the output should be.

After creating a scheduled task, it’s a good idea to run it once manually to handle any permission prompts. Then you can switch it to automatic approvals so it runs in the background—even when your computer is off, if it’s configured to run in the cloud.

Beyond level 6: Claude Code for building tools

Everything above focuses on using Claude as a powerful assistant. Claude Code is the next step: using Claude to build tools and software, even if you’re not a programmer.

Claude Code and Cowork share the same underlying engine. The difference is what they’re pointed at:

• Cowork focuses on everyday work with your files and apps.
• Claude Code focuses on writing and running code, building applications, and automating complex workflows.

You still talk to Claude Code in normal language. The “code” part describes what it produces, not who it’s for.

From design to working website

A powerful pattern is to combine Claude Design and Claude Code:

1. Use Claude Design to generate a full visual design of a website or interface based on a detailed brief.
2. Export that design and hand it to Claude Code.
3. Ask Claude Code to turn the design into a real, functioning website in a local project folder.

Claude Code will create the necessary files, wire up interactions, and even integrate with services like your email marketing platform using connectors and APIs. You can then customize, deploy, and connect it to e-commerce platforms like Shopify.

If you want a deeper dive into this side of the platform, there’s a full walkthrough in the ultimate beginner’s guide to Claude Code.

Building reusable tools that outlive the chat

One of the biggest mindset shifts is realizing that Claude Code doesn’t just do a task once—it can build a tool that keeps doing that task forever.

For example, instead of asking Claude to generate labels every time, you can:

• Ask Claude Code to build a small app that reads your inventory spreadsheet and generates a print-ready PDF of product labels.
• Have it place a shortcut on your desktop.
• Update your spreadsheet whenever you add new products, then double-click the app to generate fresh labels.

At that point, Claude isn’t in the loop anymore. You’re just running a tool on your computer, no conversation required.

Climbing the Claude ladder at your own pace

You can think of using Claude as climbing a six-level ladder:

1. Talk to Claude – Learn good prompting and refinement.
2. Give it your world – Files, images, and web research.
3. Make real things – Documents, spreadsheets, artefacts, and mini tools.
4. Make Claude remember – Projects, instructions, and personal memory across devices.
5. Connect your apps – Use connectors and skills so Claude can work inside your tools.
6. Let Claude work alone – Cowork, scheduled tasks, and eventually Claude Code building tools.

If you can handle level 1, you can grow into every other level. The core habit never changes: give clear context, define the task, specify the format, and refine the result. The only difference is how big and powerful those results become.