Give context before tasks
Do not just say "write me a caption." Say "I am a business coach for consultants. I just ran a workshop on pricing. Write me an Instagram caption sharing one takeaway." The more context, the better the output.
Every Claude concept explained simply. Ordered by what matters most for your business. No jargon. No fluff.
The difference between "meh" and "magic" is not skill. It is setup.
Every conversation starts cold. Claude does not know your business, your voice, or your audience. You get generic, one-size-fits-all answers.
Claude already knows who you are, how you talk, who you serve, and what you are working on. Every response feels tailored to you.
Claude comes in three flavors. You only need to care about the first two.
The web and mobile app at claude.ai. You type, Claude responds. Great for writing, brainstorming, research, and answering questions. This is where most people start and where you will set up your Master Prompt and Projects.
The desktop app with superpowers. Can access files on your computer, connect to your apps (Gmail, Calendar, Drive), run tasks in the background, and work while you sleep. This is where Claude becomes your assistant, not just a chatbot.
A command-line tool for software developers. It reads entire codebases and writes code across multiple files. If you are not a developer, you can completely skip this one.
Most people are stuck at Level 1. Your job is to move to Level 2 and then Level 3.
Using Claude like a search engine. One-off questions, one-off answers. Every conversation starts from scratch.
Claude knows your business, your voice, your offers, and your audience. Every conversation starts warm, not cold.
Claude runs workflows for you. Content, campaigns, follow-ups, briefings, client work. It does the work, not just answers questions.
These are the foundations. Set them up once and every conversation gets better.
Must KnowSeparate brains inside Claude. Each project has its own memory, instructions, and conversations. Create one for "My Business", one for "Content", one for "Personal". They keep everything organized so your business context does not bleed into your personal stuff.
A set of rules that tells Claude how to operate inside a specific project. It loads automatically at the start of every conversation. Claude already knows your rules, preferences, and context without you repeating yourself. In Cowork this is called the CLAUDE.md file. In Chat, it is just "Instructions" inside your project settings.
Your permanent identity across ALL of Claude. These apply everywhere, in every project, every conversation. Unlike project instructions (which only apply to one project), global instructions follow you into everything.
How Claude remembers things about you over time. It picks up patterns from your conversations and stores important details. The more you use Claude, the better it gets. You can also tell it directly to remember something. Think of it as Claude building a profile of you conversation by conversation.
Bridges between Claude and the apps you already use. Google Calendar, Gmail, Google Drive, Notion, Slack, Stripe, and 30+ more. Instead of copy-pasting information into Claude, it reaches into your tools directly and pulls what it needs, or takes action for you.
Claude can search the internet for live, up-to-date information. Useful for research, competitor analysis, trending topics, and anything recent. Just ask and it searches automatically. No special setup needed.
Visual outputs that go beyond plain text. Claude can generate dashboards, slide decks, charts, interactive tools, and more. You can view them right inside Claude and open them in your browser. This is one of the things that makes Claude feel like magic.
A folder on your computer where all your Cowork files live. Every file Claude generates (documents, scripts, dashboards) is saved here. Think of it as Claude's filing cabinet. You can access it from Finder/Explorer at any time. (Fun fact: this entire cheat sheet was built inside Marc's workspace folder — which is, honestly, his biggest organizational weakness. Ironic.)
/Archive folder — any file you think you do not need anymore, ask Claude to move it there. Review once a month and delete what you truly do not need. Way safer than deleting on the spot.You can upload documents directly into a project as "knowledge." Your offer page, your client onboarding doc, your pricing breakdown, your best blog posts. Claude reads these and uses them as reference every time you chat inside that project. This is how Claude learns the specifics of YOUR business, not just what you tell it in instructions.
Built-in presets that change how Claude writes. You can choose from styles like "Formal", "Concise", or "Explanatory" — or create your own custom style by giving Claude examples of your writing. Custom styles are powerful because Claude matches your voice without you having to explain it every time.
Once you have the essentials down, these features multiply your results.
Worth LearningPre-built workflows that tell Claude HOW to do a specific task the way you want it done. Instead of explaining how you want something done every time, you teach Claude once and it follows that same process forever. One command triggers an entire workflow. Skills live in Cowork but you can also trigger them inside Claude Chat projects.
Bundles of skills packaged together for easy sharing and installation. Instead of creating 20 skills one by one, you install a plugin and get them all at once. Anthropic has built-in plugins for marketing, legal, design, and more. You can also create and share your own.
Automations that run without you. Daily briefings, weekly reports, content calendars. You tell Claude what to do and when, and it just runs on schedule. This is where Claude starts replacing tools like Zapier. For example: a daily community scan, or a weekly email performance report.
Trigger words that start a skill with a single command. Type /morning-briefing and the entire workflow runs. Type /carousel and it starts your carousel creation process. It is just a shortcut to trigger your skills faster.
A Chrome browser extension that lets Claude interact with websites for you. It can fill out forms, click buttons, pull information from web pages, and navigate sites. Think of it as giving Claude hands inside your browser. Useful for apps that do not have a connector yet.
A fancy word for "Claude can see things." It can read images, screenshots, charts, PDFs, not just text. Upload a photo of your whiteboard notes and Claude can read them. Screenshot a competitor's landing page and ask Claude to analyze it.
You do not need these on day one. But knowing they exist means you know where you are heading.
Good to KnowThink of this as Claude's desk. Every conversation has a limited amount of space. Your instructions, memory, files, and conversation history all take up space on that desk. Currently Claude can handle about 1 million tokens (roughly 700,000 words). If your conversation gets very long, Claude may need to summarize older parts to keep working.
When you give Claude a complex problem, it can take extra time to think through it carefully before responding. Like asking someone to take their time. The answer takes longer but comes out better.
Claude can actually use your computer, not just your browser. It can open apps, click through menus, and interact with desktop software. Still early stage and slower than connectors, but it handles edge cases where no connector or Chrome extension exists.
Control Claude on your desktop computer from your phone. If you are out and need Claude to grab a file, run a task, or check something, Dispatch lets you trigger it remotely. Like a remote control for your AI assistant.
Parallel workers that handle complex tasks by splitting the work. Give Claude a big request and it breaks it into parts: one agent does research, another writes copy, another builds a dashboard. All at the same time. Claude usually triggers this automatically when a task is complex enough.
How Claude runs code behind the scenes in Cowork. When you ask it to resize images, process a CSV, or generate a report, it writes and runs code automatically. You never see or touch the code.
Tips that separate people who dabble with AI from people who run their business with it.
Do not just say "write me a caption." Say "I am a business coach for consultants. I just ran a workshop on pricing. Write me an Instagram caption sharing one takeaway." The more context, the better the output.
Paste 2-3 examples of posts, emails, or copy you love. Say "Write in this style." Claude mirrors patterns better than it follows abstract descriptions.
If the first response is 70% there, say what is wrong and ask Claude to fix it. "Make it punchier." "Remove the fluff." "Make the hook stronger." Do not start a new conversation. Build on what you have.
"Respond as if you are a copywriter who has written 500 landing pages." "Respond as if you are my business coach." This shifts the quality of reasoning Claude applies.
Create separate projects for business, content, personal, and client work. Do not dump everything into one project. Your business voice should not mix with your personal journaling.
Long instructions get diluted. Focus on: who you are, who you serve, your voice and tone, your rules, and what Claude should never do. Quality over quantity.
Claude builds memory over time but it can get stale. Once a month, check what it has stored. Remove anything outdated. Add anything important it missed.
If you do the same task more than 3 times (weekly emails, social posts, client onboarding), turn it into a skill. The time you invest once saves hours every week.
Connectors (Google Calendar, Gmail, Drive) are fast and reliable. Chrome extension is the backup. Computer use is the last resort. Work your way up, not down.
Your morning briefing. Your daily content scan. Your weekly report. Start with 1-2 scheduled tasks that save you the most time, then expand from there.
Run through this checklist to make sure your Claude is set up for maximum results.
The one thing to understand: Claude does not count messages. It counts tokens. Every time you send a new message, Claude re-reads your ENTIRE conversation history before replying. Message 30 costs roughly 31x more tokens than message 1. Once you get that, everything below makes sense.
When Claude gets it wrong, do not send "no I meant THIS" as a new message. Click edit on your original message, fix it, and regenerate. The old exchange gets replaced, not stacked on top.
Before you do, ask Claude to summarize everything, copy the summary, open a new chat, and paste it as your first message. You keep the context without the token bloat.
Three separate prompts means three full context reloads. One prompt with three tasks means one reload. The answers usually come out better too because Claude sees the full picture upfront.
If you find yourself uploading the same documents into multiple chats, upload them once into a Project instead. Every new conversation inside that Project references them without burning tokens again.
If you keep repeating the same setup — your role, your style, your preferences — save it once in Settings. Claude applies it automatically from that point on. Stop paying for re-introductions.
Haiku for grammar checks, brainstorming, formatting, quick translations. Sonnet for real work. Opus for deep thinking. Matching the model to the task frees up 50-70% of your budget for the stuff that actually needs heavy lifting.
Claude uses a rolling 5-hour window, not a midnight reset. If you use everything in one morning session, most of your daily capacity goes unused. Working in 2-3 blocks across the day gives you significantly more total usage.
PDFs and Word docs take more processing to read. Plain text or Markdown (.md, .txt) files are lean and load instantly. Before uploading a PDF, ask Claude to "Convert this to a clean Markdown file" and upload that version instead.
Paste this into any Claude chat and it will audit how you use Claude, then give you a personalized action plan to stop hitting limits. Works standalone or inside a project.
Two copy-paste prompts. One for business. One for life. Do not overthink it.
A Master Prompt is the document Claude reads FIRST every time you talk to it. It teaches Claude who you are, what you do, how you talk, and what you want help with — so you never have to re-explain yourself. You build one for your business and one for your life. Paste each into the Instructions of its own project. That is it.
How to use these: Create a project (one for business, one for life), paste the prompt below into the project chat, answer Claude's questions, then paste the final result into the project Instructions.
Paste this into your "My Business" project chat. Claude will interview you, then hand you a finished Master Prompt to save as your project Instructions.
Paste this into your "My Life" project chat. Same flow — Claude asks, you answer, you get a finished document.
The gap between generic AI and AI that actually runs your business is not skill. It is setup. Set it up right, and Claude goes from a chatbot to a business partner.