Most solopreneurs try Claude once, get a generic result, and go back to doing it themselves. The problem was setup, not AI. This guide fixes that in under 30 minutes.
You are not behind. You are not bad at this. You started with a blank tool that had no idea who you are, how you sound, or what you needed, and it gave you exactly that: blank, generic output.
Claude set up right is a different experience. It knows your name, your business, your tone, and what you want from it before you type a single word. That's what this guide builds. It takes under 30 minutes and you only do it once.
People get confused because there are several versions of Claude. Here is what each one is, when to use it, and where it shows up in the series.
Free to start. Works in any browser. No download needed. Everything in this guide runs here.
Claude on your phone. Good for capturing thoughts on the go using your phone's built-in voice-to-text.
Free download for Mac or PC. Works directly on your computer and can see your screen and read your files. More powerful than the browser for hands-on work.
A desktop tool for building custom software using plain English. No coding background needed. You describe what you want and Claude builds it.
The honest version: start free. When you hit the daily limit and it is frustrating to wait, upgrade to Pro. Don't upgrade before that.
Go through these in order. Each one builds on the last. By step seven, Claude will already know who you are before you type anything.
Go to claude.ai in your browser and sign up. Use your email address rather than a Google or Apple login, it keeps your Claude account separate and easier to manage on its own.
Start with the free plan. You can upgrade to Pro at any time from Settings > Plans. Most people try free for a day or two and upgrade once they hit the daily limit.
Custom instructions tell Claude who you are before every single conversation. Without them, Claude starts completely from scratch every time, no name, no business, no tone, nothing. This is the single most important step in the whole guide.
Where to find it: click your name or initials in the bottom left corner, then Profile > Custom Instructions. Paste in the template below and fill in your details.
My name is [your name]. I run [type of business] and I serve [describe who you work with]. When you write for me: keep the tone [describe yours, e.g., warm and direct / professional but approachable / conversational]. Use short sentences. Avoid AI-sounding words like "delve," "transformative," "unlock," or "seamless." Do not open responses with "Certainly!" or "Great question!" Do not add filler. Just answer directly. When writing content for me to publish or send, make it sound like a real person wrote it, not a company. If a draft sounds robotic, flag it and rewrite it.
Memory is different from custom instructions. Custom instructions are what you tell Claude once. Memory is what Claude notices and saves on its own as you work together over time, things like which clients you're currently working with, how you prefer to receive feedback, or what format you keep asking for. It builds a picture of how you work without you having to maintain it manually.
Where to find it: Settings > Memory. You will see two switches. "Search and reference chats" lets Claude look back at your past conversations. "Generate memory from chats" lets Claude save what it learns about you. Turn both on.
On the same Memory screen, under "You," Claude keeps two memory cards: Profile and Preferences. Profile is who you are: your name, your business, who you serve. Claude fills it in on its own as you chat, but that takes weeks. Write it yourself today and Claude starts smart instead of getting smart slowly.
How to do it: at the bottom of the Memory screen there's a message box. Paste the filled-in prompt below right there (or into any regular chat) and Claude saves it to your Profile.
Save this to your memory profile about me. My name is [your name]. I run [type of business] and I serve [who you work with]. My main offer is [what you sell and roughly what it costs]. I work [solo / with a VA / with a small team]. Right now my biggest goal is [the thing you're working toward this quarter], and my biggest bottleneck is [the task that eats your week].
Preferences is the second memory card under "You," and it covers how you want Claude to work: your tone, your format, your pet peeves. Your custom instructions from step 2 are the rules you wrote once. Preferences is the living version Claude keeps updating as it gets to know you. Keeping both aligned is what makes every chat feel consistent.
Same mechanic as your Profile: paste the filled-in prompt below into the message box at the bottom of the Memory screen, or into any chat.
Save this to your memory preferences about how I like to work. Keep your tone [warm and direct / professional but approachable / conversational]. Use short sentences and plain words. When you need information from me, ask one question at a time. When I ask for something to publish or send, write it so it sounds like a real person, and flag anything that sounds like AI. My pet peeves: [e.g. long intros before the answer, corporate words, being agreed with when I'm wrong]. Formats I like: [e.g. short bullet lists for plans, plain paragraphs for emails].
Projects are the most underused feature in Claude. Here is the difference between a regular chat and a Project chat: a regular chat has no persistent instructions, it forgets everything the moment you close it. A Project chat carries its instructions into every conversation you start inside it, automatically. Create a Project called "My Business," write two or three sentences about what you do, who you serve, and the tone you want. Every chat inside that Project already has that context before you start.
Where to find it: left sidebar, under Projects. Click "New project." Name it, then add your business description in the Project instructions field. This is separate from your custom instructions, both run at the same time.
Custom instructions and Projects tell Claude about your business. A voice file tells Claude how you actually sound. When Claude has a complete picture of your writing patterns, your spoken rhythm, and the words you do and don't use, it stops producing output that sounds like AI and starts producing output that sounds like you wrote it. Without this, you will spend time editing every single draft.
Don't have a voice file yet? Voss builds one for you in a single session. It is free and it is the tool on this site built specifically for this step. Once you have the file, upload it to your Project so every chat in that Project uses it automatically.
Find the feature you just learned about and follow the steps to set it up today. Each guide covers what it is, how to turn it on, and how to get real use out of it.
Stop guessing what Claude needs from you. One line added to any prompt turns Claude into the expert doing the interviewing. The draft you get back is built from your actual answers, not from generic assumptions.
Type what you want Claude to create, an email, a caption, a sales page, a proposal. Don't polish it. Incomplete is fine. Claude will ask for what it needs.
Before you hit send, paste this exact sentence at the end of your message. Do not change the wording.
Before you answer, ask me any questions you need to perform this task perfectly.
Claude will come back with a specific list of questions. Answer them directly. One or two sentences per question is enough. The more honest your answers, the more the output will sound like your actual business.
After you answer, Claude writes using exactly what you gave it. You will recognize your own thinking in what comes back because it came from you, not from a generic template.
A dedicated workspace that holds your instructions, uploaded files, and conversations together. Every chat you start inside a Project already knows who you are, how you work, and what you sound like. Without Projects, Claude starts from zero every single time you open a new chat.
In the left sidebar on claude.ai, click New project. Give it a name that describes what it is for. Start with one for your main business work, something like "My Business" or "Client Work."
Click Project Instructions inside the project. This is Claude's briefing for every conversation inside this folder. Write 3 to 5 sentences covering who you are, who you serve, the tone you want, and anything Claude should always keep in mind before it writes anything for you.
Drag in your brand guide, your best past newsletters, your Voss voice file, any client context you return to regularly. Claude reads everything you upload and draws on it in every conversation inside this Project.
Projects have a file storage limit. Start with your most important files and add more over time as you identify what Claude keeps needing to reference.
This one habit makes the difference. Never open a general chat for business work. Go into the Project and start it there. That is the only change between a Claude that knows you and one that does not.
Direct links between Claude and the apps where your real work lives. Once connected, Claude can read your Google Drive files, check your calendar, and pull from your Notion notes without you copying or pasting a single thing. You stop being the middleman between Claude and your own data.
Click the plus icon inside the chat bar at the bottom of any conversation. Select Connectors from the menu that appears. You will see the full list of available app integrations.
Select the integrations that make sense for your business. Google Drive, Google Calendar, and Notion are the most useful starting points for most solopreneurs. Each one walks you through a short authorization flow to give Claude read access to your data.
Once connected, you can speak to Claude as if it already has your context. "Summarize the client brief in my Drive folder." "What does my calendar look like on Thursday?" "Pull the notes from my last product launch in Notion." Claude reads the content and works with it directly in your conversation.
Before any client call, ask Claude to pull the relevant Drive files and give you a one-page summary. Before writing anything new, ask it to check your existing content for context. Connectors turn Claude from a writing tool into a system that actually knows what is happening in your business right now.
When Claude creates something standalone, a document, a tracker, a report, a formatted template, it opens in a dedicated side panel on the right of your screen instead of dropping a wall of text into the chat. You can edit directly in that panel, keep refining it in the chat without losing what was built, and download the finished version when you are done.
Go to Settings > Capabilities and make sure Artifacts is toggled on. It is usually enabled by default but worth checking if you have never seen a side panel open during a conversation.
Artifacts open automatically when Claude creates a document, tracker, dashboard, formatted report, or any reusable tool. Try asking for things like: "Build me a weekly content tracker I can reuse each Monday." "Create a client onboarding checklist I can send as a PDF." "Write my proposal as a formatted document."
You can type changes directly in the Artifact panel on the right, or ask Claude in the main chat to update it. Both work. The panel refreshes in real time and your conversation stays intact alongside it so you never lose your thread.
Click the download or copy icon at the top of the Artifact panel to save what was built. Documents export cleanly for sharing with clients. Interactive tools like trackers export as files you can reuse or hand off.
A reusable workflow you define once and trigger with a single slash command. Every time you need that process, you type / and its name and Claude runs the entire thing, with your exact standards, your exact tone, and your exact format, without you writing a single word of explanation.
In Claude, go to Customize > Skills. This is where you build and manage your saved processes.
Tell Claude exactly what the skill does. Be precise about the output you want. Not "write captions" but "audit my caption for filler words, cut it to under 150 characters, and rewrite it in my voice without using AI-sounding phrases like 'dive in' or 'game-changer.'" The more specific the description, the more consistent the skill performs every time.
Claude will ask questions about how you want the skill to run. Answer with your actual preferences, your tone, your format rules, the things you always end up correcting by hand. This is where those standards get locked in permanently so you never have to repeat them again.
Give the skill a short, clear name. In any chat, type / followed by that name to run the full process instantly. No context. No explanation. Just the result.
The desktop version of Claude. It runs directly on your Mac or PC and can see your screen, read files from your computer, and take actions locally, things the browser version cannot do. If you need Claude to work with what is actually on your machine, this is the version that does it.
Go to claude.ai and download the desktop app for your Mac or PC. Install it the same way as any other application. Sign in with your existing Claude account and your Projects and settings carry over automatically.
On first launch, Claude Co-work will ask for permission to access your screen and your local files. Grant both. Without them, the desktop app functions like the browser version and cannot see or interact with what is on your computer.
Open the app and start a new chat. Tell Claude to "look at what is on my screen" or "read the files in my downloads folder." It will describe what it sees and wait for your instruction before taking any action.
Co-work is most useful when your work involves actual files on your machine. "Organize these invoices into folders by month and year." "Read this contract and pull out every date, deadline, and dollar amount." "Look at my screen and tell me what needs my attention first."
A feature inside Claude Co-work that turns your phone into a remote control for your desktop Claude. Send a message from anywhere and your computer does the work while you are out of the office. You stay present wherever you are and the task still gets done.
Open Claude Co-work on your Mac or PC. Go to Settings and turn on both Keep computer awake and Dispatch. Your computer must stay powered on and connected to the internet for Dispatch to receive and run commands while you are away.
Open the Claude mobile app on your phone and find the Dispatch option in the menu. Follow the prompts to link your phone to your active desktop session.
Type your instruction from your phone exactly as you would at your desk. "Find the signed contract on my desktop and pull the renewal date." "Check my downloads folder for any PDF invoices from this week and summarize them." Your desktop Claude runs the task and sends the result back to your phone.
A dedicated mode inside Claude for visual and brand work. Give it your logo and your exact colors and it builds a proper design system around them, typography, color usage rules, spacing, and templates. Instead of making design decisions from scratch every time, you start with a finished, consistent foundation.
Look for Design in the mode or tool selector inside Claude. Select it to switch into this workspace. The interface shifts to support visual and layout-focused work.
Give Claude your logo file and your brand colors. Hex codes work best, for example, your primary green is #1A5C52. The more precise your input, the more accurate and consistent everything it generates will be.
"Build a full design system from these assets with color usage rules, font pairings, and spacing guidelines." "Create a 5-slide Instagram carousel template using my brand." "Design a simple one-page sales sheet layout for my next offer." Claude generates visual mockups and structured design specs you can take directly into Canva or any other tool.
A tool for building custom software using plain English. If you have ever needed a specific tracker, dashboard, booking system, or client portal and could not find an app that worked the way your business actually works, Claude Code builds it for you. You describe what you need. It builds the working tool. No technical background required at any step.
Claude Code runs as a separate tool alongside the Claude desktop app. You will need to install it on your Mac or PC. Go to claude.ai and follow the Claude Code setup instructions for your operating system. The setup takes about 10 minutes.
Type /plan and describe what you want to build in plain English. Be specific about what the tool needs to do. "A weekly tracker that shows my active clients, their current project status, and next due date, with a way to mark each item complete and filter by client." Claude maps out the full build and shows you the plan before writing a single line.
Read what Claude proposes. It will ask clarifying questions before it builds, how you want to input data, what the output should look like, what should happen when something is marked complete. Answer in plain language. This is how you stay in full control without needing to understand anything technical.
Claude builds the working tool and you test it. If something needs adjusting, describe the change the way you would explain it to a person. "Move the status column to the left." "Add a notes field under each client row." It updates. You never write code at any step.
Two separate settings that work together. Memory saves what Claude notices about you over time so you do not have to re-explain yourself in every new conversation. Privacy controls whether Anthropic uses your conversations to train future AI models. You want Memory on and data training off.
Go to Settings > Memory and toggle it on. From this point, Claude will notice and save facts about you across separate conversations, things like your active clients, your working preferences, how you like feedback delivered, and recurring tasks you handle each week.
Click Manage Memory to see the full list of what Claude has stored about you. Read through it. Delete anything that is wrong, outdated, or too personal to keep there. Claude relies on this list actively, so accuracy matters more than volume.
Go to Settings > Privacy and turn off Help improve Claude. This stops Anthropic from using your conversations as training data for future AI development. Your business notes, client details, draft copy, and private conversations stay private.
Turning off data training does not affect how Claude performs for you in any way. It only controls whether Anthropic can use your conversations to train its models. Your experience stays exactly the same either way.
Once a month, open Manage Memory and remove anything stale or no longer true about your business. Clients change. Projects end. Your preferences evolve. Keeping the list current means Claude stays accurate instead of operating on outdated information.
A way of working where Claude manages multiple specialized assistants at once, each focused on a single job, while you direct the whole operation. You stop doing tasks one piece at a time. You start reviewing finished work and deciding what happens next. This is how you move from being the busiest person in your business to being its director.
Start a new Project in Claude and name it your Operations Center or something similar. Write Project Instructions that establish Claude in this workspace as a coordinator, someone who manages a team of specialists and brings you finished output rather than asking you to do it step by step.
Inside the Project, tell Claude to set up named specialists for the areas of your business that need the most ongoing support. One for writing and content. One for research and competitor analysis. One for client communication. Give each one a specific job, a specific tone, and clear rules for how it operates. Claude keeps each role consistent.
Hand the coordinator a high-level goal rather than a task list. "Plan the full launch of my new offer and give me a content calendar, an outreach sequence, and a summary of what needs to go live on which day." Claude breaks the goal down internally, distributes the work, and brings you a finished result to review.
Your job is to read what comes back, approve what is right, and adjust what is not. You are the director. Everything else gets handled before it reaches you.
You don't need to learn a prompting system. You need to know four things. Most bad results come from skipping one of them.
Pick the one that costs you the most time right now. Copy it, fill in the orange fields, and paste it into Claude. Each one is ready to use.
Paste in the emails you have been putting off. Claude writes replies in your tone so you edit instead of write from scratch.
Here are [X] emails I need to respond to. Write a reply for each one that matches my tone: [describe it, e.g., warm and direct, short sentences, no corporate language]. Keep each reply under 5 sentences. Do not open any reply with "I hope this email finds you well." Email 1: [paste email here] Email 2: [paste email here]
Write your thoughts rough and messy. Claude shapes them into something you can actually post.
Here is what I want to say about [topic]. It's rough, that's fine: [paste your thoughts here] Turn this into one short social caption that sounds like a real person wrote it. No hashtags. No "As a [profession]" opener. No "Excited to share." Just say the thing I'm trying to say, in the simplest way possible.
Paste your full task list and let Claude help you figure out what actually needs your attention.
Here is everything on my plate this week:
[paste your list or brain dump here]
Help me figure out:
1. What actually needs my full attention this week
2. What I can batch together
3. What can wait or be skipped entirely
Keep it short. I want a prioritized list, not a motivational speech.
Upload an email, a post, a blog entry, or a transcript. Claude reshapes it into a new format without adding ideas that aren't yours.
Here is a [blog post / email / caption / recording transcript] I already have: [paste your content here] Turn this into [three short social captions / a short video script / a bullet-point summary / a short email]. Keep my voice. Do not add new ideas, just reshape what is already here into the new format.
Give Claude the context and get back a short prep sheet: what to ask, what to address early, and what you want to walk away with.
I have a [discovery call / client check-in / proposal meeting] with [brief description of who this person is]. Here is what I know about them and what we have discussed so far: [paste any notes, emails, or background here] Help me prepare: 1. What to ask them (3 questions max) 2. What to address before they bring it up 3. What I want to walk away with Keep it short. One page or less.
In a long conversation, Claude can only see a certain amount of what was said before. Think of it like this: it is reading the last several pages of the chat, not the whole thing. When the conversation gets long enough, the beginning starts to fall outside what it can see.
You will notice it when Claude contradicts something it said earlier, gives an answer that ignores context you already provided, or seems to lose track of the task. This is not a bug. It is how the tool works.
The Project instructions carry over automatically into every new chat. The conversation resets to a clean state. You keep all your context and lose all the noise. A practical rule: when a chat has gone past 15 to 20 back-and-forth exchanges, start a new one in the same Project. You won't lose anything that matters.
If you want to carry something specific from the old chat into the new one, paste a short summary at the top: "We have been working on [X]. Here's where we left off: [2–3 sentences]. Continue from there."
Every one of these is fixable. Knowing them in advance saves you the frustration of quitting after one bad result.
This is not a course. There is no homework. Three windows, one task each.
You will see advice online that says: upload your emails and texts and AI will learn your voice. That approach gets you something. Claude picks up vocabulary, sentence length, and rough tone from a stack of examples.
What it misses: the difference between how you write and how you actually speak. The patterns you use without realizing. The way your voice shifts when you're selling versus teaching versus delivering hard news to a client. The rhythm that is yours out loud but doesn't show up on the page.
A pile of examples shows Claude what you've said. Voss shows it how you think. One session. A finished voice file that documents your real patterns, not a surface read of old content, and works in any AI platform, not just Claude.