Setting up your Claude account

Claude is Anthropic's AI assistant, available at claude.ai. Getting your account configured correctly in the first session saves hours of friction later. This reference guide covers every setup decision you'll face — plan selection, interface options, organization features, and the settings that actually affect how Claude responds to you. Bookmark this page. You'll return to it when you upgrade your plan, add a team member, or wonder why Claude's response style shifted. Everything here is current as of Claude's 2024-2025 product lineup, including the claude.ai web app, iOS and Android apps, and API access.

7 Things to Know Before You Click 'Sign Up'

1. Claude is made by Anthropic, not OpenAI — it is a direct competitor to ChatGPT, not a version of it.
2. The free tier gives you access to Claude 3.5 Haiku (a fast, capable model) with daily message limits that reset every 24 hours.
3. Claude Pro costs $20/month — identical pricing to ChatGPT Plus — and unlocks Claude 3.5 Sonnet and Claude 3 Opus with significantly higher usage limits.
4. You sign up with a Google account or email address. No credit card required for the free tier.
5. Claude has no persistent memory by default — it forgets everything when you close a conversation, unless you use the Projects feature (Pro only).
6. The API is separate from claude.ai — developers pay per token through console.anthropic.com, while this lesson focuses on the consumer web app.
7. Claude is available in most countries, but Anthropic restricts access in a small number of jurisdictions due to legal and compliance requirements.

Creating Your Account

Go to claude.ai and click 'Sign up'. You can authenticate with Google (fastest), or use any email address with a password. Anthropic sends a verification email — check your spam folder if it doesn't arrive within two minutes. Once verified, you land directly in the chat interface. There's no onboarding wizard, no profile questionnaire. Claude is ready to use immediately. The interface is intentionally minimal: a text input at the bottom, your conversation history on the left sidebar, and a model selector at the top of each new conversation.

Your account settings live behind the icon in the bottom-left corner of the sidebar. This is where you'll find billing, notification preferences, and the option to set a display name. Claude uses your display name in conversations if you introduce yourself, though it doesn't automatically pull it into every response. The settings panel is sparse by design — Anthropic deliberately limits the number of toggles and configuration options compared to competitors like ChatGPT, which has a more complex settings architecture including custom instructions, memory management, and plugin configuration.

• Account verification takes under 2 minutes via email link
• Google OAuth login skips the email verification step entirely
• Display name is optional but useful if you want Claude to address you by name
• You can change your email address in account settings post-signup
• Two-factor authentication (2FA) is available and recommended for Pro accounts
• Deleting your account permanently removes all conversation history — no recovery option

Plan Comparison: Free vs. Pro vs. Team

Understanding Claude's Models

Every conversation you start with Claude runs on a specific model — and which model you're using has a direct impact on response quality, speed, and capability. The model selector appears at the top of each new conversation window. Free users see Claude 3.5 Haiku as their primary option. Pro users can switch between Haiku, Sonnet, and Opus. Think of these as a speed-to-power spectrum: Haiku is fast and efficient, Sonnet balances speed and intelligence, and Opus is the most capable but slowest model Anthropic offers.

For most professional tasks — writing, analysis, summarization, brainstorming — Claude 3.5 Sonnet is the right choice. It's what Anthropic positions as their flagship for everyday use. Opus is worth switching to for genuinely complex reasoning tasks: multi-step analysis, code architecture decisions, or processing a 150-page document where you need deep comprehension. Haiku shines when you need fast, repeated outputs — running the same prompt against multiple inputs, quick rewrites, or simple Q&A. Choosing the wrong model wastes either time (Opus for simple tasks) or quality (Haiku for complex ones).

1. Open a new conversation by clicking the pencil/compose icon in the sidebar
2. Look for the model name displayed at the top center of the chat window
3. Click the model name to open the selector dropdown
4. Choose your model based on task complexity — Sonnet for most work, Opus for deep analysis
5. Your model choice applies only to that conversation — each new conversation defaults to your last selection

The Projects Feature (Pro and Team Only)

How Projects Work

Projects solve Claude's biggest limitation for professional users: the lack of persistent memory. Without Projects, every conversation starts from zero — Claude has no idea who you are, what your company does, or what style you prefer. Projects let you store a set of instructions and uploaded documents that Claude reads at the start of every conversation within that project. This is the closest Claude currently gets to 'remembering' you. A Project functions like a persistent briefing document that travels with every conversation you have inside it.

Create a Project by clicking 'New Project' in the sidebar. Give it a name (e.g., 'Q1 Marketing Copy' or 'Client Analysis — Acme Corp'), then add your project instructions in the text field provided. These instructions can include your role, your company context, preferred output formats, tone guidelines, and any recurring constraints. You can also upload reference files — brand guidelines, style guides, product specs, org charts — up to the file size limits. Every conversation you start inside that Project inherits this context automatically, without you re-typing it.

Quick Setup Task: Your First Project

Account Setup Cheat Sheet

• Sign up at claude.ai — Google login is fastest, no credit card needed for free tier
• Free plan: Claude 3.5 Haiku, ~30-40 messages/day, no Projects
• Pro plan: $20/month, Claude 3.5 Sonnet + Opus, Projects, priority access
• Team plan: $25/user/month, minimum 5 seats, adds shared workspaces and admin controls
• Model selector is per-conversation — choose Sonnet for most work, Opus for complex analysis
• Context window is 200K tokens (≈150K words) across all models and plans
• Projects store persistent instructions + files — Pro and Team only
• Projects do NOT auto-update from conversations — maintain them manually
• Enable 2FA in account settings for security, especially on Pro accounts
• Account deletion is permanent — export anything important before deleting

Key Takeaways

1. Claude and ChatGPT are direct competitors with identical Pro pricing ($20/month) — Claude's main differentiator at this tier is the 200K token context window and the Projects feature.
2. Model choice matters: Haiku for speed, Sonnet for most professional tasks, Opus for complex reasoning. The wrong choice costs you either time or quality.
3. The free tier is genuinely usable for occasional tasks, but the lack of Projects makes it inefficient for any recurring professional workflow.
4. Projects are the most important setup decision for professional users — configure one correctly and you eliminate the need to re-explain your context in every conversation.
5. Claude's interface is intentionally minimal — don't look for extensive settings menus. The power is in how you prompt, not in configuration toggles.

Navigating the Claude Interface

With your account created and your plan chosen, you're looking at the Claude interface for the first time. It's deliberately minimal — a text input, a conversation thread, and a sidebar. That simplicity is intentional. Unlike some AI tools that bury you in buttons and modes, Claude is designed around conversation. But minimal doesn't mean featureless. Knowing where everything lives and what it actually does saves you from wasting time clicking around when you should be prompting.

The Sidebar: Projects and Conversation History

The left sidebar holds two things: your conversation history and Projects. Every chat you start is saved automatically and searchable by keyword. Claude Pro users can also create Projects — persistent workspaces where Claude retains context across multiple sessions. A Project for "Q3 Marketing Campaign" remembers your brand voice guidelines, your target audience notes, and your previous outputs every time you return. Free users get conversation history but no Projects, which means each new chat starts from scratch without any carry-over memory.

• Conversation history is searchable — use keywords from your original prompt to find old chats
• Projects (Pro only) maintain persistent memory and custom instructions across sessions
• Each new conversation outside a Project starts with zero memory of previous chats
• You can rename conversations manually — do this for anything you'll want to find later
• Archived chats are not deleted; they're hidden from the main sidebar view
• Projects support uploaded files that stay available for the lifetime of the project

Claude Plans at a Glance

Understanding Context Windows

The 200,000-token context window on Claude Pro is one of its biggest practical advantages. A token is roughly 0.75 words in English, so 200,000 tokens handles approximately 150,000 words — a full novel, or an entire company's quarterly report. This matters because Claude can hold your entire conversation, plus any uploaded documents, in working memory simultaneously. You can paste in a 50-page contract and ask questions about specific clauses without Claude losing track of the beginning by the time it reaches the end.

Context windows reset with every new conversation. Whatever you built up across a long chat — the background you provided, the decisions made, the style Claude adopted — disappears when you click "New Chat." This is why Projects matter for ongoing work. Think of a Project as a persistent context container. Think of a standard conversation as a whiteboard that gets erased when you leave the room. Choosing the wrong one for a task costs you significant setup time.

File Uploads: What Claude Can Actually Read

Claude accepts more file types than most users realize. You can upload PDFs, Word documents, spreadsheets, plain text files, code files, and images. For images, Claude performs genuine visual analysis — it reads charts, interprets diagrams, and extracts text from screenshots. This isn't decorative multimodal capability; it's practically useful for analyzing competitor screenshots, reviewing presentation slides, or pulling data from a chart someone sent as a PNG instead of a proper dataset.

Your First Real Prompt

Custom Instructions and System Prompts

Inside a Project, you can set custom instructions — a block of text Claude reads before every conversation in that Project. This is where you define standing context: your role, your company, your preferred output format, your audience. Done well, custom instructions eliminate the repetitive setup text you'd otherwise paste at the start of every prompt. A consultant might set instructions that define their client's industry, the engagement's goals, and the required deliverable format. Every prompt in that Project inherits that context automatically.

Custom instructions work best when they're specific and behavioral, not vague and aspirational. "Be helpful and professional" does almost nothing — Claude already behaves that way. "Always respond in bullet points unless I ask for prose. Use British English. When reviewing documents, flag legal and financial risks separately from operational risks" — that's actionable. Instructions that specify format, audience, vocabulary, and scope produce consistent, on-brand outputs without extra prompting effort every session.

Quick Setup Task: Create Your First Project

Interface Quick-Reference Cheat Sheet

• New Chat button — starts a blank conversation with no memory of previous chats
• New Project — creates a persistent workspace with custom instructions and file storage
• Paperclip icon — attach files (PDF, images, spreadsheets, code, text)
• Rename conversation — click the chat title in the sidebar to edit
• Retry button — regenerates Claude's last response with a different output
• Copy response — copies Claude's full reply as plain text to your clipboard
• Thumbs up / down — sends feedback to Anthropic; doesn't affect your conversation
• Model selector (Pro) — switch between Claude 3.5 Sonnet and Claude 3 Opus mid-conversation
• Project settings — access custom instructions, uploaded project files, and sharing options (Team plan)

Making Claude Work for You: Settings, Habits, and First Wins

Your account is live. Now the difference between casual users and power users comes down to three things: knowing which settings actually matter, building a consistent prompting habit, and understanding Claude's guardrails so you never waste a conversation hitting a wall. This section covers the practical configuration layer most users skip entirely — and shows you how to get a real, keeper output from Claude in your first session.

Profile and Memory Settings That Actually Matter

Claude.ai's profile settings let you tell Claude who you are before you type a single prompt. Fill in your role, your industry, and how you prefer responses formatted. Claude uses this context in every conversation — a marketing director who sets her profile gets tighter, business-oriented answers than someone with a blank profile asking the identical question. This isn't cosmetic. It compresses the setup work you'd otherwise do manually at the start of each chat.

• Profile name and role: Claude uses this to calibrate formality and assumed knowledge level
• Response style preference: choose concise, balanced, or detailed as your default
• Projects (Claude.ai Pro): persistent memory spaces that retain context across sessions
• Custom instructions: plain-English rules Claude applies to every response you receive
• Language preference: set your working language once, not per conversation

Understanding Claude's Boundaries

Claude will decline certain requests — not randomly, but according to Anthropic's published usage policies. Knowing the pattern saves frustration. Claude won't generate content that facilitates real harm, produce deceptive material designed to mislead specific people, or write detailed instructions for dangerous activities. These refusals are consistent and intentional. What surprises new users is that Claude will push back on ambiguous requests too, often asking a clarifying question rather than guessing wrong.

When Claude declines, it tells you why — briefly. That explanation is useful. It usually signals how to reframe the prompt legitimately. Asking Claude to 'write a phishing email' fails. Asking Claude to 'write a security awareness training example showing phishing tactics employees should recognize' works, because the intent is defensive, not deceptive. Context and stated purpose change outcomes significantly.

Your First Real Prompt Session

The fastest way to calibrate Claude to your work is to give it a task you'd normally spend 30-45 minutes on. Not a test. Not 'write me a poem.' Give it a real deliverable — a draft email to a difficult stakeholder, a summary of a document you paste in, a structured agenda for a meeting next week. Real inputs produce outputs you can actually evaluate and keep. That feedback loop, repeated over your first five sessions, builds your prompting instincts faster than any tutorial.

Quick-Reference Cheat Sheet

• Free tier: claude.ai, no credit card, limited to Claude 3.5 Haiku with usage caps
• Pro tier: $20/month, full Claude 3.5 Sonnet and Opus access, Projects, higher limits
• Custom Instructions: set once in profile, apply to every conversation automatically
• Projects (Pro): persistent memory spaces — use one per major client or workstream
• Claude won't save conversation history if you disable it in Data Controls
• Refusals come with explanations — use that explanation to reframe your prompt
• Never paste confidential or regulated data into consumer Claude tiers
• Paste real work into Claude from session one — fake prompts produce fake calibration
• Claude API (separate from claude.ai) is available at console.anthropic.com for developers

Key Takeaways

1. Your profile and custom instructions are the highest-leverage settings in your account — fill them in before your first real session
2. Claude's refusals follow a consistent policy logic; reframing with context and purpose resolves most friction
3. The Free tier is a fully functional starting point; upgrade to Pro when you hit usage limits or need Projects
4. Your first five real-work sessions with Claude calibrate your prompting instincts faster than any passive learning
5. Data hygiene matters from day one — anonymize sensitive content before it enters any consumer AI tool