The Complete Claude Guide.

Most people use Claude like a smarter Google search.

Type a question. Read the answer. Close the tab.

And then they say AI is overhyped.

It's not. They just never got past the prompt box.

Claude has projects that remember everything about your work. Skills that execute your repeatable processes automatically. Connectors that pull your actual tools and data into every conversation. Research that goes to 169 sources and synthesizes everything in minutes. Artifacts that produce interactive dashboards, presentations, and live documents, not text you have to copy into something else.

This guide covers every feature. More importantly, it covers how to connect them into a system that actually works.

Claude has one of the strongest free plans of any AI model. You can get genuine value from it.

That said: if you use AI regularly, the $20/month plan is worth it. It makes a meaningful difference in output quality, speed, and the features you can access.

Most people never need to go beyond the $20 plan unless they start using Claude Cowork or Claude Code extensively, and those are separate tools beyond what this guide covers.

Before anything else: go to Settings and look at the fields that say "tell Claude about yourself" and "personal preferences Claude should consider."

Everyone recommends filling these out. Don't.

Here's why: these instructions apply to every single conversation, across every topic, no matter the context. If you use Claude across a wide range of tasks, writing, research, analysis, coding, a single instruction set doesn't fit all of them. You'll end up with weird, off-tone responses in the wrong situations.

Leave these completely blank.

Set up context in projects instead. Projects let you give Claude specific instructions scoped to a specific type of work. The right instructions apply at the right time. It's a dramatically better system, and I'll cover it in detail below.

Before getting into the advanced features, the foundation matters. A bad prompt produces generic results. The same question, asked better, produces something actually useful.

The three parts every good prompt needs: Instructions, Context, Constraints.

Instructions: define exactly what you want Claude to do. The task and the action.

Context: everything Claude needs to know to do it well. Your role, your situation, your objective, relevant background. Don't be stingy here. A full context dump is fine.

Constraints: the rules. Tone, style, length, output format, what to avoid.

The bad version:

Generic answers. Not wrong, just useless.

The better version:

Same instruction. Completely different output.

One more tip: at the end of your prompt, add this line:

I call this a context interview. Claude will ask for the information it knows it's missing, things you wouldn't have thought to include. Answer the questions. Then ask for the actual output.

The first response after a context interview is almost always usable without further iteration.

Web search is available under the + button. Leave it enabled by default.

Claude doesn't always use it automatically, it will when the need is obvious, but for tasks like critiquing a script or reviewing a proposal, it won't think to search unless you tell it to.

The technique I use constantly: ground first, ask second.

Instead of asking a question directly, run a separate prompt first to give Claude current knowledge on the topic:

Then ask your actual question. The quality of every answer that follows is dramatically better because Claude is working from current, relevant knowledge instead of its training data.

Example: before asking Claude to help with a product launch strategy, run this first:

Now when you ask for help with your launch, it responds using those frameworks. Not generic marketing advice. A specific strategy built on knowledge you gave it.

Drag files directly into the chat or use the + button. PDFs, images, spreadsheets, contracts, competitor landing pages, Claude reads, summarizes, extracts, and analyzes.

A few real examples of what this looks like:

Screenshot analysis: I dropped in a screenshot of my last 12 YouTube video thumbnails and asked: "Analyze the title and thumbnail combinations. Identify what's working visually and in messaging, what isn't, and any patterns you notice. Be direct, honest feedback, not flattery."

It flagged that I was restating the title in the thumbnail instead of extending it. It identified that list-style titles underperformed because they feel like homework. Specific, actionable, correct.

CSV analysis: I exported 90 days of YouTube analytics as a CSV and asked: "Analyze my performance trends and visualize the key metrics. What is the data telling me about what's working?"

It analyzed the data and created a full interactive analytics dashboard, without me asking for a dashboard. Claude decides the best way to present information. Sometimes that's a table. Sometimes it builds an entire visual interface.

One honest limitation: Claude can read and analyze images but cannot generate them. Every other major model, ChatGPT, Gemini, can generate images. Claude can't. If visual generation matters to your workflow, that's a genuine gap.

Artifacts are standalone interactive outputs that Claude creates in a dedicated side panel alongside your conversation.

When you ask for something substantial, a dashboard, a document, a presentation, a landing page, a flowchart, Claude builds it in a separate window. You can iterate on it without losing it in the conversation scroll. The document stays visible while you give feedback.

Some things I've built as artifacts in one prompt each:

If Claude doesn't automatically open an artifact when you expect one, just add: "Create this as an artifact."

The most useful application for writing: ask Claude to create any document as an artifact. Your document appears on the right. The chat is on the left. You give feedback, see changes in real time, and never lose your place.

Under the + button, there's a Research option. This is fundamentally different from web search.

When you click Research and ask a question, Claude doesn't run a single search. It operates agentically, it makes a plan, runs multiple searches that build on each other, adapts based on what it finds, and explores different angles of the question on its own.

On a recent research task: 5 minutes, 169 sources, a fully synthesized document with citations.

Use research for anything that would normally require hours of reading across multiple sources. Competitive analysis, market research, technical deep dives, anything where you need a real synthesis, not a quick answer.

Projects are self-contained workspaces with their own memory, chat history, knowledge base, and custom instructions.

Each project is a dedicated environment for a specific type of work. The custom instructions you set only apply inside that project. The files you upload are always available in that project. Every conversation inside inherits all of that context automatically.

No re-explaining. No re-uploading. No re-writing context from scratch.

How to set one up:

What goes in the instructions:

What goes in the files:

One project for writing. One for research. One for client work. One for your own content. Each one gets the right context for the right job.

The limitation projects don't solve: Projects remember things. They don't know how to do things. That's what skills are for.

Projects are persistent reusable knowledge. Skills are persistent reusable processes.

A skill is a saved workflow. You build it once, Claude learns exactly how to complete a specific task in a specific way, and from that point forward, it invokes the skill automatically whenever you work on relevant tasks.

There are two types:

Anthropic skills: built and maintained by Anthropic, available to all paid users, invoked automatically when relevant.

Custom skills: workflows you build for your specific needs. These are the ones that compound.

How to build one:

I wrote the guide I wish I had, free resources, real examples, and my personal skill library you can copy.

Grab it: https://www.simplifyingcomplexity.tech/p/claude-skills-full-guide

Connectors let Claude access your actual apps. Not just answer questions about them, access them. Read information, pull data, take actions on your behalf.

Google Drive, Google Calendar, Slack, Asana, and dozens more.

What this looks like in practice:

I have Granola connected, it captures my meeting notes automatically. After any strategy call, I ask Claude to pull in the notes and start building on the ideas. The meeting context is already in the conversation. I don't have to explain what we discussed.

I also use Gamma for presentations. Claude can draft the content, Gamma can build the visual. Connected.

Whatever tools you use every day, connect them. Claude stops being a chatbot and becomes the command center where everything comes together.

Opus: the most powerful model. Best for complex tasks that require deep reasoning. Slower, uses more tokens (Claude's processing currency, you have a rolling 5-hour window and a weekly cap depending on your plan).

Sonnet with Extended Thinking: my default for almost everything. Fast enough for regular use, smart enough for serious work. Extended Thinking lets Claude work through complex problems more carefully before responding.

Haiku: fast, but noticeably weaker. I never use it.

My rule: Sonnet with Extended Thinking for daily work. Opus when I'm building something that has to be right.

Claude Cowork is Claude Code for people who don't code. And if you're reading this thinking "I've opened Cowork before, it looked like a chat," you're not using it. Not really.

I built a full beginner-friendly course that covers the complete setup, every feature, and the exact workflows that actually work.

Grab it here: simplifyingcomplexity.tech/p/claude-cowork-masterclass

Everything up to this point can be used in isolation. The real jump happens when you connect them.

Here's what the full system looks like:

Projects hold your context, brand guidelines, style docs, past examples, instructions. Always available. Never re-explained.

Skills hold your processes, the exact way you want specific tasks done. Always invoked. Never re-explained.

Connectors pull in your tools, your meetings, your docs, your calendar, your apps. Always accessible.

Artifacts hold your outputs, living documents you can iterate on without losing them in the scroll.

Research handles anything that requires synthesis across sources.

The result: you open a project, your context is there. You start a task, the right skill fires. You need data from a meeting, Claude pulls it in. You want a deliverable, it builds it as an artifact.

That's the difference between using Claude as a chatbot and using it as a system.

If this changed how you think about Claude, send it to one person still just typing questions into the prompt box. That's all I'm asking.

See you in the next one :)