Claude is an AI assistant from Anthropic, the second most used program of its kind after ChatGPT. This guide is for someone who has never used it and may not feel confident with computers: what it is, how it differs from ChatGPT, how to create an account and ask the first question, what Claude does differently (documents in the side panel, projects, longer thinking mode), what the plans and prices are, which models to choose, and what Claude Code is, which does not chat but works on your files. Inside are ready commands to copy with one click.
Last verified: September 6, 2026. Informational material. The appearance of screens, model names, prices, and limits change every few months; numbers come from Anthropic documentation as of the verification date. Current pricing: claude.com/pricing.
In Brief
- Claude works like ChatGPT: you write in Polish, and it responds. It differs in the producer (Anthropic instead of OpenAI), response style, document handling, and the fact that it has a version for developers, Claude Code, which edits files and runs commands on your computer by itself.
- The account is free, but you must be 18 years old (Anthropic rules). This is different from ChatGPT, where the threshold is 13 years with parental consent.
- The free plan allows conversations with the Sonnet 5 model, web searching, memory between conversations, and apps for phone and computer; it has usage limits counted in five-hour sessions (pricing, about limits). Pro costs 20 USD per month and adds more usage, more models, and Claude Code.
- Models: Haiku 4.5 for quick tasks, Sonnet 5 for everyday use (default), Opus 5 for difficult tasks, Fable 5.1 the most powerful, with Max plans (which models in which plan). For starters, leave the default.
- Claude Code is not a chat. It is a tool that receives a task (“fix this bug”, “add a form”) and reads files, writes code, runs commands, and asks for permission before any significant change. For those who program or want to start; requires a Pro plan or higher (documentation).
- The same rules apply everywhere: verify important facts at the source, do not enter personal data and passwords, and in Claude Code, read what it asks for permission for before clicking “yes”.
What Claude Is, in Simple Terms
Claude is a program that understands and writes text, just like ChatGPT. You type a question or command in plain language, in Polish, and it responds in full sentences: explains, translates, summarizes, writes emails and documents, calculates, organizes plans, reads attached documents and images. It is made by Anthropic, founded by former OpenAI employees, who have emphasized from the beginning that the assistant should be cautious, fair, and predictable.
It works on the same principle as other large language models: it has learned from a vast amount of text, what words follow which, and constructs a response word by word. Thus, it has the same flaw: it does not fact-check, it only assembles probable text, and sometimes confidently writes something untrue. Anthropic recommends verifying important information just like OpenAI. If you are reading this guide for the first time, it is worth first reviewing our ChatGPT Beginner's Guide: sections on what a prompt is, how tokens work, and how to write good commands apply to Claude equally.
Claude vs. ChatGPT: What Is Really Different
| Feature | Claude | ChatGPT |
|---|---|---|
| Producer | Anthropic | OpenAI |
| Minimum Age | 18 years | 13 years, under 18 with parental consent |
| Free Plan | Sonnet 5, searching, memory, apps; limits in five-hour sessions | Text conversations without limits with safeguards; limits on images, files, voice; ads in the USA |
| Base Paid Plan Price | Pro 20 USD (17 USD with annual payment) | Plus 20 USD, cheaper Go 8 USD |
| Document Handling | Results appear in the side panel as a separate document (artifact) that can be edited and downloaded | Canvas, separate window next to the conversation |
| Image Generation | Does not generate images; creates charts, diagrams, and pages in the panel | Yes, also video in higher plans |
| Developer Version | Claude Code: terminal, VS Code, app, browser, phone | Codex |
| Style | Usually longer, more cautious responses; often asks when it does not know | Shorter, more direct; more “all-in-one” features |
Which to choose? Many people have both and use ChatGPT for quick questions and images, and Claude for longer texts, documents, and code. Both are free to start, so the simplest way is to try both for a week.
What You Need
- Device with internet: phone, tablet, or computer. On a computer, a browser is enough.
- Email address or Google or Apple account to log in.
- Must be 18 years old. Anthropic requires adulthood and can ask for age verification if something indicates it (about age verification).
- No payment card needed for the free version.
Step by Step: How to Open Claude and Create an Account
On a Computer, in a Browser
- Go to claude.ai
Type in the address bar
claude.aiand press Enter. This is the only real chat address; the company page is claude.com. - Choose a login method
“Continue with Google”, “Continue with Apple” or email address. With email, you will receive a code or confirmation link in your inbox; there is no password to create.
- Provide your name and date of birth
At the first login. If under 18, the account will not be created.
- You will see the chat window
A box in the middle of the screen with the question, how can I help. This is where you write.
On Phone and Computer as an App
- App Store
App Store or Google Play, type “Claude”. The publisher should be named “Anthropic”. There are many fakes with similar names.
- Computer App
Mac and Windows, available for download from claude.com/download. Not needed to start.
- Log in with the same account
Conversations from phone and computer are shared.
First Screen: What Is Where
| Element | Where | What for |
|---|---|---|
| Writing box | In the middle, then at the bottom of the screen | This is where you write. Enter sends; Shift+Enter creates a new line. |
| “+” button | Next to the writing box | Attaching files and images, enabling web search, longer thinking mode, other tools (Anthropic help). |
| Slash “/” | Typed in the box | List of commands and shortcuts. |
| Model name | Next to the writing box or at the top | Which model is responding. For starters, leave the default. |
| Side panel on the right | Appears automatically | This is where Claude shows documents, tables, charts, and code it creates (artifacts). They can be edited and downloaded. |
| Conversation list and projects | On the left | Conversation history; projects are folders for one matter. |
| Settings | Clicking on the initials in the corner | Language, memory, data, plan, log out. |
First Question: Try Now
Each command below is in a box with a “Copy prompt” button. Click, paste it into the box in Claude, and press Enter.
Explain to me in three simple sentences what you are, how you differ from ChatGPT, and how you can help me in everyday life. Write in Polish, without difficult words.
Prepare as a separate document a simple checklist for a week-long trip to Poland for a family with two children: documents, medications, electronics, clothes, matters to settle before departure. In Polish, in bullet points, with space for checking off.
Write a short, polite email in English to my child's school informing them that they will be absent on Friday due to a doctor's appointment, and request to send the assignments. Below, provide a translation in Polish.
The conversation continues. After the response, write in the same window one of the following:
Shorter, by half.
Simpler, as if explaining to a child.
More formally.
Make it a table.
Save it as a separate document in the panel.
Translate into English.
How do you know that? Search the web and provide links.
What Claude Does Differently Than ChatGPT
Artifacts – when you request a document, table, chart, simple page, or code, Claude creates them in the panel on the right instead of in the conversation content. You can edit them (“change the header to...”), switch versions, and download. Just write “make this a separate document”.
Projects – a folder for one matter (e.g., “apartment”, “business”, “learning English”), to which you attach files and enter permanent instructions; each conversation in the project knows them (Anthropic help). In the free plan, the number of projects is limited; in Pro, there is no limit (pricing).
Longer Thinking – in settings or the “+” menu, you adjust how much effort Claude puts into a response; for difficult tasks (multiple conditions, numbers, document analysis), enable a higher level, the response will be slower and more careful (Anthropic help).
Long Documents – Claude handles large files well: contracts, reports, several PDFs at once (about attaching documents). This is one of the main reasons people choose it for work.
Memory and Searching in Conversations – Claude remembers context between conversations as a list of topics that you can review and delete; searching old conversations is in paid plans (Anthropic help).
Voice – voice conversation is available in all plans, works best in the phone app (Anthropic help).
What It Lacks – Claude does not generate images or videos. If you need that, use ChatGPT or Gemini.
How to Write to Claude: The Same Four Elements, Two More Habits
A good prompt for Claude has a role, context, task, and format, just like in the ChatGPT guide. Two habits that work particularly well in Claude:
- Long, organized commands. Claude handles several-point instructions well. If you have a complex task, list it in bullet points instead of one sentence; do not be afraid of “too much”.
- “Ask me first”. Claude is happy to ask questions when you allow it. The sentence “before you answer, ask what you need” yields better results than guessing.
You are [role, e.g., an experienced advisor / teacher / editor]. My situation: [two or three sentences about you and why you are doing this]. Task: [what to do, in one sentence]. The answer should be: [length], in the form of [list / table / document in the panel / email], in Polish, [in simple words / formally]. Before you answer, ask what you do not know about my situation. Do not make up facts; if you are unsure, write that.
Ten Prompts to Start
I am attaching a letter in English. Explain to me in Polish, in simple words: who is writing it, what they want from me, by when, and what will happen if I do not respond. Then list step by step what I need to do. Do not write a response to the letter yet.
I am attaching two offers [e.g., car insurance / phone plan]. Make a table in the panel: what is in one and what is not in the other, annual costs, hidden fees, cancellation conditions. At the end, list questions I should ask both companies before I choose. Do not say which to choose.
You are my English teacher. My level: [beginner / intermediate]. Teach me today 5 useful sentences [in a situation, e.g., at the doctor’s]. Provide the sentence, pronunciation written in Polish, and translation. Then give me one exercise and wait for my answer before you evaluate.
Make a repayment plan for my debts: [list: name, balance, interest rate, minimum payment]. I have [amount] monthly for this. Show in a table two ways (first the most expensive, then the smallest), how many months each will take and how much interest I will pay. Note that this is an approximate calculation, not financial advice.
Explain to me [topic, e.g., how Social Security works / what a revolving credit is] as if I were 12 years old, with one comparison from life. Then ask me three checking questions and wait for the answers.
Summarize the attached document in five points in Polish. Separately list: dates and deadlines, amounts, obligations on my side, and fragments that you do not understand or that are unclear. Do not add anything that is not in the document; provide the page for each point.
In this project, you help me with the matter of [e.g., renting an apartment in Chicago]. Respond in Polish, simply and briefly. Leave English terms in parentheses. Use attached files as a source of facts; if something is not in them, say so. For legal and financial matters, remind to confirm with a specialist.
I am pasting a message I received. Assess whether it could be a scam: list warning signs and say what not to do. Do not assume it is real because it looks official.
MESSAGE:
[paste without your data]
I am a person who [e.g., recently arrived in the USA / runs a small business / cares for an elderly parent]. Ask me five questions to understand my situation, and then suggest ten things you could realistically help me with in the next month. Start with questions and wait for answers.
Make a simple bar chart of my monthly expenses in the panel: [category: amount, several items]. Below the chart, one sentence about where the most goes. Without embellishments.
Plans and Prices
Prices in USD, as of the verification date, according to Anthropic pricing and help on choosing a plan. All plans have usage limits counted in five-hour sessions; when you write a lot or attach large files, the limit runs out faster (about limits).
| Plan | Monthly Price | What It Offers | For Whom |
|---|---|---|---|
| Free | 0 | Conversations with Sonnet 5, web searching, memory, apps, document and code creation in the panel; limited usage; no Claude Code | Anyone starting out |
| Pro | 20 (17 with annual payment) | More usage, more models (including Opus 5), unlimited projects, Claude Code, and other work tools; Fable model for additional credits (about Pro) | Daily use, document work, beginner programmers |
| Max | 100 or 200 | Five or twenty times more usage than Pro, higher limits, Fable 5.1 included up to half the weekly limit, priority during peak hours (about Max) | People working with Claude all day, intensive Claude Code |
| Team | 20–25 per person | For teams, shared billing, Claude Code | Companies |
| Enterprise | Individually | Administrative control, security | Large organizations |
When Is It Worth Paying? When the free version's session limit runs out in the middle of work, when you need Opus 5 for difficult analyses, when you want to have many projects, or when you want to try Claude Code. The plan can be changed and canceled in account settings.
Which Model to Choose
Models are “engines” that differ in speed, depth, and limit consumption. As of the verification date, the family looks like this (model overview, Anthropic guide on choosing):
The fastest and most economical. Simple questions, short summaries, quick translations.
For everyday use: writing, analysis, documents, code, research. Default in the free and Pro plan. Sufficient for most tasks.
For tasks that really require deep thinking. Opus from the Pro plan; Fable 5.1 in Max included, in Pro for credits. Consume the limit faster.
Advice for beginners: leave Sonnet 5 and enable longer thinking for difficult tasks. Change the model only when you know why.
Claude Code: What It Is and Who It’s For
Claude chats. Claude Code works. You give it a task in plain language (“add a contact form to my page”, “find out why this script crashes”, “organize these files and describe what you did”), and it reads files on your computer, writes and modifies code, runs commands, checks the results, and asks for permission before any significant change (documentation, product page). It is a tool for people who write programs or want to start; it requires a Pro plan or higher (Anthropic help).
| Feature | Claude (chat) | Claude Code |
|---|---|---|
| How It Looks | Chat window in the browser or app | Terminal window, code editor extension, computer app, claude.ai/code page, or phone |
| What It Does | Writes a response; documents in the panel | Reads and modifies files, runs commands, creates commits and pull requests, connects to tools |
| Who Decides | You copy the result wherever you want | Claude suggests, you approve; changes go directly to the files |
| For Whom | Everyone | Programmers and those learning to program; advanced users for automation |
| Plan | Free and above | Pro and above (or developer account with pay-per-use) |
Where It Works
The same engine is available in several places and remembers project settings regardless of where you run it (documentation):
- Terminal (command line) on Mac, Linux, and Windows: full version, for those who are not afraid of a black window.
- Code Editor: extension for VS Code (also Cursor) and plugin for JetBrains; changes are visible as colored differences in files.
- Computer App (Mac, Windows): “Code” tab in the Claude app, without installing anything in the terminal; the easiest start for a beginner; requires a paid plan.
- Browser and Phone: claude.ai/code works on projects from GitHub in the cloud; a task can be assigned from the phone and returned for results.
Installation in Terminal, Step by Step
If you have never used a terminal, start with the computer app from the previous point. If you want the terminal version, the following commands come from Anthropic documentation; you paste them into the terminal (on Mac, the program “Terminal”, on Windows “PowerShell”) and press Enter.
curl -fsSL https://claude.ai/install.sh | bash
irm https://claude.ai/install.ps1 | iex
cd path/to/my/project
claude
Upon first launch, Claude Code will ask you to log in with your Claude account in the browser. Then in the same window, you write tasks in plain language, in Polish too. Before it changes or runs anything, it will show what it intends to do and ask for permission.
First Tasks for Claude Code
It is safest to start with tasks that only read, not change. The following are entered already inside Claude Code, after launching in the project folder.
Read this project and explain to me in Polish, in simple words, what it does, what files it consists of, and where to start if I want to change something. Do not change anything.
I want to [e.g., add a “Contact” page with a form / correct typos in texts / add a README file describing the project]. First, show me the plan: which files you will change and why. Wait for my approval before changing anything. After the changes, list what exactly you did.
When I run [program name or command], I get this error:
[paste error content]
Find the cause, explain it to me in simple words, and suggest a fix. Before you implement it, show the difference in files and ask for permission.
Create a CLAUDE.md file in this project with rules to follow in every session: respond in Polish; show the plan before each change; do not delete files without asking; after changes, run tests if there are any; describe commits briefly in English. Show me the content before saving.
The CLAUDE.md file in the project folder is read at the beginning of each session; it is a place for your permanent rules (documentation). It is also advisable to have a copy of the project (preferably in a version control system like git) so that any change can be reverted.
- Read what it asks. The request for permission to run a command is the moment you decide. Do not click “yes” blindly; if you do not understand the command, ask: “explain what this command does”.
- Work on a copy. A folder with a backup or a git repository. Claude Code can change many files at once; reverting without a copy can be difficult.
- Do not run in folders with personal documents or with passwords. Run in a specific project folder.
Security and Privacy
- Do not enter Social Security numbers, card and account numbers, passwords, or other people's data. They are never needed for a good answer.
- Set whether your conversations can train the models. In privacy settings, there is a separate switch (Anthropic article).
- Check important things. Dates, amounts, regulations, doses: ask for sources with web search enabled and click on them.
- Fakes. The real Claude is claude.ai, the publisher's apps from Anthropic, and claude.com/download. Pages requesting payment for “access” are not Anthropic.
- Age 18+. An account for a child is not possible; for learning with younger ones, it is better to work together on an adult's account with their involvement.
About what you can safely do with an AI assistant in official matters, we write in the guide ChatGPT and US Agencies; the rules apply to Claude the same way.
Common Beginner Mistakes
- Asking like to a search engine. A full sentence with context gives a better answer.
- New conversation for each correction. Write corrections in the same conversation.
- Forgetting about searching. With current prices, regulations, and events, enable web search in the “+” menu or write “search the web”.
- Attaching a document without asking. The file alone does nothing; write what it should do with it.
- Changing the model to “the best” for everything. Opus and Fable consume the limit faster; Sonnet 5 is sufficient for most tasks.
- Claude Code on the entire disk. Run in the project folder, on a copy, and read requests for permission.
- Believing the first answer. Ask for sources and for listing what it is unsure about.
Frequently Asked Questions
Is Claude free?
Yes, the free plan exists and is sufficient for daily conversations. It has a usage limit in five-hour sessions; with large files and long conversations, it runs out faster. Pro costs 20 USD per month.
Does it speak Polish?
Yes, fluently, in writing and voice. For texts to American agencies and companies, ask for an English version with Polish translation.
Do I need Claude Code if I don’t program?
No. Claude in chat does everything a non-programmer needs. Claude Code makes sense when you want to build or fix programs and websites or automate work on files on your computer.
How does Claude Code differ from Codex in ChatGPT?
They are tools of the same kind: agents that work on code. They differ in producer, model, and operational details. If you have a Pro plan in Claude, you have Claude Code included; Codex details are described by OpenAI.
Why won’t Claude make me an image?
Because it does not have image generation capabilities. It makes charts, diagrams, and simple pages in the panel. For images, use ChatGPT or Gemini.
What happens to my conversations?
They are saved on your account; you can delete them. Whether they are used to train models depends on the switch in privacy settings, which you can turn off.
I ran out of limit. What now?
Wait for the end of the five-hour session or upgrade to a higher plan. The limit runs out faster with large files, long conversations, and stronger models; a new conversation for a new topic saves the limit.
Fact-Check Summary
- Definitely true: Claude is made by Anthropic; it works on claude.ai, in phone and computer apps; an account requires 18 years; a free plan exists and has usage limits in sessions; Claude Code is available in the terminal, editors, app, and browser and requires a Pro plan or higher; Claude does not generate images. Everything with links to Anthropic documentation.
- Probably true as of the verification date: Pro prices 20 USD, Max 100 and 200 USD, Team 20–25 USD per person; Sonnet 5 as the default model; Fable 5.1 in Max included, in Pro for credits. This changes most often; check the pricing.
- What is uncertain: the exact size of session limits for each plan and how many projects the free plan has; Anthropic changes them without notice.
- Common myth: “Claude Code is for everyone.” It is a tool for working on code and files; a non-programmer only needs chat.
Sources
| Source | Type | Credibility |
|---|---|---|
| Anthropic – Claude Pricing | Official | High |
| Anthropic – Choosing a Plan | Official | High |
| Anthropic – Pro Plan | Official | High |
| Anthropic – Max Plan | Official | High |
| Anthropic – Fable Models in Plans | Official | High |
| Anthropic – Usage Limits | Official | High |
| Anthropic – Minimum Age | Official | High |
| Anthropic – Age Verification | Official | High |
| Anthropic – Getting Started | Official | High |
| Anthropic – Download Apps | Official | High |
| Anthropic – Projects | Official | High |
| Anthropic – Uploading Documents | Official | High |
| Anthropic – Web Search | Official | High |
| Anthropic – Memory and Searching in Conversations | Official | High |
| Anthropic – Voice Mode | Official | High |
| Anthropic – Model Overview | Official | High |
| Anthropic – How to Choose a Model | Official | High |
| Anthropic – Claude Code, Overview and Installation | Official | High |
| Anthropic – Claude Code, CLAUDE.md File | Official | High |
| Anthropic – Claude Code Page | Official | High |
| Anthropic – Claude Code in Pro and Max Plans | Official | High |
| Anthropic – Is My Data Used for Model Training | Official | High |
Comments (0)
No comments yet. Be the first!