Everything you need to know about using Claude — explained in plain English. No tech degree required.
Getting up and running takes about 2 minutes. Here's your step-by-step:
claude.ai in the address bar.Think of these as three different "modes" for working with Claude. They all use the same AI brain behind the scenes, but they're designed for different situations — kind of like how you might use your phone for texting, email, or a phone call depending on the situation.
| Mode | What It Is | Best For | Real-World Example |
|---|---|---|---|
| Chat | The regular chat window — like texting back and forth. You go to claude.ai in your browser and type. | Quick questions, brainstorming, writing emails, summarizing documents, general help | "Draft a follow-up email to the client from yesterday's meeting" |
| Code | A tool that runs inside your computer's terminal (the black text window). It can read, write, and edit files on your computer. | Building tools, editing code, automating tasks, working with files and folders on your machine | "Read this sales spreadsheet and tell me which accounts are past due" |
| Cowork | Claude works alongside you in the background while you do other things. It watches what you're working on and helps proactively. | Pair-working sessions where you want Claude watching over your shoulder and chiming in with suggestions | You're editing a financial model and Claude flags a formula error before you even ask |
Opus and Sonnet are two different "models" of Claude — think of them like trim levels on a car. Same brand, different capabilities and price points.
| Sonnet | Opus | |
|---|---|---|
| Speed | Faster responses | Slower, more deliberate |
| Smarts | Very capable — great for most tasks | The most intelligent — best for complex reasoning |
| Cost | Cheaper per use | More expensive per use |
| Best For | Everyday tasks: emails, summaries, quick questions, basic analysis | Hard stuff: complex financial analysis, nuanced writing, multi-step problem solving |
Think of Sonnet as a sharp, experienced associate — fast, handles most things perfectly, great value. Think of Opus as a senior partner — takes a bit more time, costs more, but when the problem is really complicated, you want them on it.
This is the single most important thing you can learn. The quality of your results depends almost entirely on how you ask. The good news? It's not complicated. Think of it like giving instructions to a smart new hire on their first day — they're brilliant, but they know nothing about your situation.
1. Be Specific, Not Vague
"Help me with an email."
"Write a professional but friendly email to a client letting them know their project deliverables will be ready two days ahead of schedule. Include that we'd love to schedule a walkthrough call at their convenience."
2. Give Context — Explain the "Why"
Tell Claude who you are, who the audience is, and what the end goal is. This helps it make better judgment calls.
3. Describe the Format You Want
Don't make Claude guess how you want the answer. Tell it:
4. Give an Example When It Matters
If you have a specific format or style in mind, showing Claude an example is worth a thousand words of explanation.
"Format the client summary like this example:
Acme Corp — Q1 2026 Summary
Revenue: $142K | Growth: +18% YoY | Satisfaction: 9.2/10
Status: Strong. Contract renewal expected in Q3.
Now do the same for the other three clients using this data..."
5. Tell Claude It's OK to Ask Questions
If you're not sure you've covered everything, add: "If anything is unclear, ask me before proceeding." This prevents Claude from guessing wrong and saves you time.
| Before (Vague) | After (Specific) |
|---|---|
| "Summarize this contract." | "Summarize the key business terms of this contract in bullet points: payment terms, deliverables, timeline, termination clauses, and any unusual provisions." |
| "Write a report." | "Write a one-page performance summary for Q1 covering revenue, client retention, and key wins. Format it for our all-hands meeting." |
| "Help me with this spreadsheet." | "Look at the sales data in this Excel file and tell me: which accounts are more than 30 days past due, what's the total outstanding balance, and who are the repeat late payers?" |
Here are practical, real-world ways Claude can save you time — starting with the easiest and most impactful:
A Skill is like a shortcut or a pre-programmed recipe for Claude Code. Instead of typing out a long, detailed prompt every time, a skill bundles it all up into one simple command.
Think of it like speed dial on your phone. Instead of dialing a 10-digit number every time, you just press one button. Skills work the same way — instead of typing a paragraph of instructions, you type a short command and Claude already knows what to do.
Skills are triggered by typing a slash command inside Claude Code. For example:
/commit — tells Claude to save your code changes with a nice description/review-pr — tells Claude to review a set of code changes and give feedbackSkills can be built-in (they come with Claude Code) or custom-made. For example, you could create a custom skill like /weekly-report that automatically pulls your data and drafts a summary every Monday.
A token is how Claude measures the amount of text it reads and writes. It's the unit of measurement behind the scenes — and it's what determines usage costs when using the API.
Roughly speaking:
If you're using the API (the behind-the-scenes connection that powers tools like Claude Code), you're charged based on how many tokens Claude reads and writes. Think of it like a meter running:
If you upload a 20-page contract and ask for a summary, that uses more tokens (and costs more) than asking "What's 2+2?"
There are two common ways to pay for Claude:
Either way, it's far cheaper than hiring a person to do the same work.
Claude is incredibly capable, but it's not perfect. Knowing what it can't do is just as important as knowing what it can — it'll save you from trusting the wrong thing.
| Limitation | What This Means for You |
|---|---|
| It can "hallucinate" | Claude can sometimes make things up and present them confidently as fact. This is especially risky with specific numbers, legal citations, or industry data. Always verify critical facts. |
| No real-time information | Claude doesn't browse the internet by default. It won't know today's stock prices, yesterday's news, or current market conditions unless you provide that data. |
| No access to your systems | Claude Chat can't log into your CRM, check your email, or pull from your company's databases. You need to provide the data (copy/paste or upload files). |
| Math can be shaky | Claude is a language model, not a calculator. For critical financial calculations (profit margins, loan amortization, tax projections), double-check the math or use a spreadsheet. |
| It can't remember across conversations | Each new conversation starts fresh. Claude doesn't remember what you talked about yesterday. If you need context, provide it at the start of each chat. |
This is a great question, and the short answer is: Anthropic takes this very seriously. Here's how your data is protected:
| What You Might Worry About | What Actually Happens |
|---|---|
| "Will other people see my conversations?" | No. Your conversations are private to your account. Other users, other companies, and the public cannot see them. |
| "Is Claude learning from my data and sharing it with others?" | No. When you use the API (which is what Claude Code uses), Anthropic does not use your inputs or outputs to train their models. Your data stays yours. |
| "Is my data being stored forever?" | Anthropic may temporarily store data for safety and abuse monitoring (typically up to 30 days), but it is not used for training and is automatically deleted. |
| "Can Anthropic employees read my stuff?" | Only in rare cases for safety reviews or if required by law. They have strict internal access controls. Day-to-day, no one at Anthropic is reading your conversations. |
Even with these protections, it's smart to follow a few common-sense rules:
No question is too basic. The best way to learn is to open Claude and start experimenting. If something doesn't work the way you expect, just ask Claude: "Why did you do it that way?" — it's happy to explain.