FAQ · Start here

Claude FAQ
for Beginners.

The most common questions a first-time Claude user asks, with short answers. Read top to bottom if you're new. Skip to the band that matters if you're past the basics.
INFLXD. FAQ · Vol. 01

The basics.

If you've never opened Claude before, start here. Three quick framings before anything else.

What is Claude?

Claude is an AI assistant built by Anthropic. You chat with it to write, summarize, analyze documents, reason through problems, and code. Anthropic is an AI safety company, which is why Claude tends to flag uncertainty and push back instead of confidently bluffing.

How was Claude trained to behave the way it does?

Most AI assistants are trained by humans rating thousands of responses as good or bad. Anthropic added a step on top called Constitutional AI: Claude is trained against a written list of principles (a "constitution") and taught to critique its own answers against them.

The practical effect: Claude is more willing to acknowledge uncertainty and more careful about pushback than models trained on pure human feedback. You don't need the details to use it, but if you've ever wondered why Claude sometimes says "I'm not sure," this is why.

How is Claude different from ChatGPT?

Both are general-purpose AI assistants and both cost $20 a month at the consumer tier. The practical difference: Claude writes and reasons like a careful colleague, ChatGPT does more multimedia in one app (image generation, video, voice). For writing, analysis, and coding, most people find Claude better. For "make me a picture of a cat," ChatGPT.

Getting in.

Sign-up, plans, where to use Claude, and which model to pick. The four decisions you make before typing anything.

How do I sign up and start using Claude?

Go to claude.ai and sign up with email or Google SSO. No credit card needed for the free plan. You can start chatting in the browser immediately. At INFLXD, use your @inflxd.com Google account so you land on the company workspace.

Is Claude free? What do I get?

Yes, there's a free tier. You get roughly 15 to 40 messages per rolling 5-hour window (the exact number depends on message length and server load), access to current models, web search, basic Artifacts, voice mode on mobile, and limited Projects. Free is fine for trying Claude out. You'll hit the limit fast during real work.

What does Claude Pro get me, and is it worth $20/month?

Pro is $20/month, or $17/month if you pay annually. You get roughly 5x the usage of free during peak hours, plus Projects, the Research feature, Google Workspace integration, Claude Code access, and Cowork. There's a weekly cap on top of the 5-hour window. For anyone using Claude daily for work, Pro pays for itself in a week.

Should I use the desktop app or the website?

Either works. The website runs in any browser, no install. The desktop app (claude.com/download, macOS 11+ or Windows 10+) adds three things you can't get in the browser: Quick Entry (a keyboard shortcut to summon Claude from any app), one-click extensions for tools like Slack and Google Drive, and Cowork mode. If you're going to use Claude every day, install the desktop app.

Which Claude model should I pick, Opus, Sonnet, or Haiku?

Three models. Sonnet is the everyday default, fast and capable. Opus is the heavyweight for the hardest reasoning and longest documents. Haiku is the lightweight for quick lookups and high-volume tasks.

Model Use when Speed Cost
Haiku 4.5 Fast, simple lookups and short summaries Fastest Lowest
Sonnet 4.6 Default for almost everything Fast Medium
Opus 4.7 Hardest reasoning, longest documents, complex code Slowest Highest

If you don't think about it, Sonnet is correct.

Using it day-to-day.

The four things you'll actually do once you're inside: prompt, upload, search, and ask for visuals.

How do I write a good prompt?

Be specific about what you want. Give Claude the context it needs (who you are, who the output is for, what good looks like). Iterate, your first prompt rarely produces the final answer. For complex tasks, break them into steps. You don't need fancy "prompt engineering" tricks. Clarity beats cleverness.

What file types can I upload, and how big can they be?

Claude takes PDFs, Word docs (DOCX), Excel (XLSX), PowerPoint (PPTX), plain text (TXT, CSV, MD, HTML), and standard images (JPEG, PNG, GIF, WebP). The cap is 30MB per file and 20 files per conversation. PDFs under 100 pages get full visual analysis including charts and images. Longer PDFs are processed text-only. Drag and drop into the chat, or use the "+" button.

Can Claude search the web?

Yes. Click the "+" or slider icon in the chat input and toggle on Web search. Claude will fetch live results when you ask about anything time-sensitive. On Team or Enterprise plans, an admin has to enable it first. If you want to force the issue, just say "search the web" in your prompt.

Can Claude generate images?

Not natively. Claude has no built-in text-to-image model, so it can't make a photo or illustration the way DALL-E or Midjourney does. It can produce SVG vector graphics, diagrams (Mermaid), and small interactive components via Artifacts, which covers most business needs. For real image generation, use ChatGPT, Midjourney, or connect an external image model.

More than a chatbot.

Features that turn Claude from "chat tool" into "system you build on." Skim these once. They unlock real compounding the second time around.

What are Projects and when should I use them?

A Project is a persistent workspace where Claude remembers your custom instructions and references uploaded files across every chat inside it. Use one when you have recurring work with the same context: a client account, an ongoing report, a course you teach. Upload reference docs once, write a short brief, and every conversation in the Project inherits both. Projects are best on a paid plan.

What are Artifacts?

Artifacts are a side panel that opens when Claude produces something substantial and reusable: a document, a chunk of code, an SVG, a diagram, a small interactive app. You can edit them, share a public link without recipients needing a Claude account, and iterate by highlighting specific text and asking Claude to rewrite just that part. Available on every plan, including free.

Does Claude remember me between conversations?

Yes, if you've enabled memory. Claude auto-generates a synthesis of your past chats roughly every 24 hours and uses it as context in new conversations. You can also tell it directly what to remember. Each Project has its own separate memory. Review, edit, or disable memory in Settings > Capabilities.

What are Skills and Plugins?

Two ways to give Claude new capabilities without changing how you chat.

A Skill is a packaged set of instructions Claude loads when it sees a matching trigger phrase. Think saved playbook. You build (or download) a skill once, install it by dragging the .skill file into Claude Desktop, and after that whenever you say the trigger phrase, Claude follows the playbook automatically. Examples: an "EOD report" skill that drafts your end-of-day update in your manager's voice, or a "glossary" skill that extracts new terms from any transcript.

A Plugin is a bundle that can include skills, tool connectors (Slack, Drive, Linear, etc.), and other capabilities, all installed together for a specific role or workflow. Think of it as the starter pack version of a skill: pick one matching your job and you get a curated set of capabilities in one install.

You don't need either to get value from Claude. But once you've used it for a few weeks, the patterns you find yourself repeating are exactly the ones worth packaging into a skill.

Trust and safety.

Three questions every careful user eventually asks. Worth a minute each.

What's Claude's knowledge cutoff?

It varies by model. Sonnet 4.6 has a reliable knowledge cutoff around August 2025, with some training data going through January 2026. For anything more recent (current news, this week's prices, today's headlines), turn on web search rather than trusting Claude's memory.

Is it safe to put confidential work info into Claude?

It depends on the plan. On Free, Anthropic may use your conversations for product improvement and safety research, so don't paste anything sensitive. On Pro and Max, conversations aren't used for training by default, and you can verify your opt-out in Privacy Settings. For truly confidential work (client data, financials, regulated information) your company needs a Team or Enterprise plan with a Data Processing Agreement. At INFLXD, check with the AI enablement lead before pasting client data.

Can I trust what Claude tells me? Does it make things up?

Sometimes, yes. All current AI models can hallucinate, meaning they produce confident-sounding text that isn't true. Claude is generally more cautious than most and tends to flag uncertainty, but you should still verify factual claims, citations, and numbers before using them in client work. Rule of thumb: trust Claude for first drafts, structure, and thinking-partner work. Verify before publishing.

The wider Claude.

Beyond claude.ai chat, three other surfaces. You probably don't need all three, but it helps to know what they are.

What is Cowork mode?

Cowork is a research preview feature in the Claude desktop app. It lets Claude work agentically on your files and apps: read, edit, and create things inside folders you give it access to, then run multi-step tasks on its own while looping you in. Think "Claude Code, but for knowledge work." Available on Pro and Max for macOS. Currently in research preview, so expect rough edges. Claude can only touch folders and connectors you explicitly grant.

What is Claude Code and do I need it?

Claude Code is Anthropic's agentic coding tool. It runs in your terminal, VS Code, or as a desktop app, reads your whole codebase, and edits files and runs commands on your approval. Included in Pro and above. If you don't write code, you don't need it. If you do, it's the most popular AI coding tool in 2026 and worth installing. Get it at code.claude.com.

What is the Claude API and when would I use it?

The API lets developers send messages to Claude programmatically and build it into their own apps or scripts. Billed separately from Pro. You create an account at console.anthropic.com, add at least $5 of credits, and generate an API key (it starts with sk-ant-). Keep the key secret. API usage is governed by commercial terms, which means inputs are never used for training, stricter privacy than the consumer chat. Most INFLXD teammates don't need this. It's only relevant if you're building automations.

Where this came from.

Sources for each answer above, in case you want to verify or read further.