Verdict: If you pay $20 a month for Claude and only use the chat box, you are using roughly one-eighth of what you're paying for. The other seven capabilities live behind a single settings menu that most people never open. Setting them up — in the right order — takes under an hour and changes Claude from a search engine into an assistant that knows you, sees your work, and runs tasks while you're not watching. This guide walks through all seven, verified against Anthropic's own documentation and current research on how people actually use AI.
Last verified: 2026-07-30 · Format: how-to setup guide · Primary keyword: how to set up Claude · Best for: professionals and small-business owners on a Claude Pro ($20/mo) or Free plan · Volatile facts: feature names and limits change with Anthropic updates — re-check before relying on specifics.
The real reason most people use Claude at 10% of its capacity
Only 5.5% of knowledge workers meet the bar for AI proficiency, according to a 2026 study by the AI training company Section, which surveyed 5,026 U.S. knowledge workers and tested them — rather than asking them to self-report — on what they know, what they use, and what they can build. The largest group, 73.5%, are what Section calls "Experimenters": fluent enough to use AI daily, but only for one-off tasks with nothing repeatable. Another 20.9% barely engage at all. The most common use, at 61%, is "looking things up" — a $20/month search engine. (Section, The AI Proficiency Report, 2026)
The gap is not intelligence, age, or technical skill. It's onboarding. You don't use Claude — you move into it, the same way you onboard a new hire: tell it who you are, show it where things live, and then let it work without you watching. The seven settings below do exactly that, organized into three moves that build on each other.
Block 1: Give Claude a memory (Settings 1–3)
The first three settings stop Claude from meeting you as a stranger every morning. All three live in Settings (click your profile in the bottom-left corner of claude.ai, then Settings).
1. How to write Claude custom instructions (the highest-leverage 60 seconds)
Near the top of the Settings menu you'll find a field called Instructions for Claude. Most people never see it, so it sits empty and Claude opens every conversation guessing who it's speaking to. Write to it the way you'd brief a capable new hire on day one:
- What you do and who you serve
- How you want to be spoken to ("get to the point, challenge me, tell me when I'm wrong")
- The language you use and the language you avoid
Keep it general — this is what stays true across everything you do. The specifics (a particular client, a project, rules for one job) belong in the next settings, not here. Whatever you write here, Claude carries into every conversation from now on. Anthropic's own documentation confirms this is designed as standing context that applies to every chat. (Anthropic, Claude Help Center)
2. How to turn on Claude memory (so it stops forgetting between chats)
The second setting is the one most people confuse with the first. Instructions is what you tell Claude about yourself. Memory is what Claude learns on its own and keeps.
In the same Settings menu, open Capabilities → Memory. You'll see two switches. Turn both on:
- The first lets Claude look back across your past conversations
- The second lets it carry what matters forward
With both on, the client you explained on Tuesday is still understood on Friday. Your pricing, constraints, and the details you're tired of repeating stay known. You remain in control — open the panel anytime and read exactly what it holds, then delete anything you'd rather it didn't keep.
3. How to use Claude Projects (a sealed room for each client or job)
Projects are where the specifics belong — the ones you kept out of your Instructions. This feature is free: even a free Claude account gets five projects. A Pro plan ($20/month) gives you unlimited projects. (Anthropic, Claude Help Center — Projects)
In the left sidebar, open Projects and create a new one. Name it. You've just built a room. Each room holds its own memory, its own documents, its own rules, sealed off from the rest. Into your best client's room, put their pricing, your proposal template, the last few things you sent them — then close the door. Every conversation you start inside that room begins already knowing all of it.
Block 2: Give Claude access to your work (Settings 4–5)
Everything so far required you to type information to Claude. These two settings let Claude go and see for itself.
4. How to connect Claude to Gmail, Google Drive, Calendar, and more
Go to Settings → Capabilities → Connectors. You'll find a catalog with Gmail, Google Calendar, Google Drive, Slack, Notion, and 50+ others. Anthropic's documentation confirms these are first-party integrations available across all Claude products, requiring no setup beyond authentication. (Claude.ai Documentation — Connectors overview)
Connect the tools you live in. Approve once and you're done. What this buys you: you can now say "review the last month of email with this client and tell me what I've promised them that I haven't delivered." That's not a chat prompt — that's a task you'd hand an employee.
Important limits to know:
- Gmail is draft-only: Claude can create drafts but cannot send emails on your behalf. (Claude Help Center — Google Workspace connectors)
- Microsoft 365 is read-only: Claude can search your Outlook mail, calendar, and Teams but cannot create, change, or send.
- Connectors only work inside a conversation you start — there are no event triggers. If you want Claude to act on its own, that's Block 3.
5. How to turn on Claude Skills (so it delivers files, not descriptions)
In the same Capabilities menu, open Skills. A skill is a job description that Claude reads before it starts working. There are several pre-built — Excel, PowerPoint, PDF — and you can create custom ones.
The difference is not small. With the Excel skill on, you ask for a spreadsheet and receive an actual spreadsheet with working formulas, tabs, and formatting — a file you can send, not a description of the work. That's the line between an answer and a deliverable.
Block 3: Let Claude work without you (Settings 6–7)
Everything so far makes Claude better when you ask. This last move means you stop having to ask.
6. How to set up Claude Co-work (point Claude at your real files)
In Co-work, instead of talking to Claude in a window, you point it at a real folder on your computer. Claude works on your actual files directly. You choose the folder, give it the job. This is useful on its own — but it becomes genuinely transformative when you add the next setting.
7. How to schedule Claude tasks (the setting that runs your week)
This is the one most people don't know exists. In Co-work, you can set up Scheduled Tasks: put in a job once, schedule it, then step away. Every Monday at 7 a.m., before you're awake, Claude works through your numbers, business updates, and emails, and prepares an update for you. You don't have to ask for it.
This is the difference between the person who uses Claude and the person who has moved into it. In Section's study, only 16% of knowledge workers had ever used an agentic tool that runs on its own, and only 13% had set up an automated workflow. (Section, The AI Proficiency Report, 2026) Microsoft's 2026 Work Trend Index, which surveyed 20,000 AI-using knowledge workers across 10 markets, found a similar pattern: broad adoption, but narrow depth. (Microsoft, 2026 Work Trend Index)
The 7 Claude settings at a glance
| # | Setting | Where to find it | What it does | Free or paid? |
|---|---|---|---|---|
| 1 | Instructions for Claude | Settings → top of page | Standing brief about who you are, carried into every chat | All plans |
| 2 | Memory | Settings → Capabilities → Memory | Claude remembers across conversations; you can read and delete what it holds | All plans |
| 3 | Projects | Left sidebar → Projects | Sealed rooms per client/job with own memory, docs, and rules | Free: 5 projects · Pro: unlimited |
| 4 | Connectors | Settings → Capabilities → Connectors | Claude reads and acts in Gmail, Drive, Calendar, Slack, 50+ tools | All plans |
| 5 | Skills | Settings → Capabilities → Skills | Claude delivers actual files (spreadsheets, decks, PDFs) not just text | All plans |
| 6 | Co-work | Co-work (left sidebar) | Claude works on real folders on your computer | Pro and above |
| 7 | Scheduled Tasks | Co-work → Scheduled Tasks | Claude runs jobs on a schedule while you're away | Pro and above |
Pricing/limits change often — last checked 2026-07-30. Verify against Anthropic's pricing page before relying on specifics.
What this means for you
If you run a small business or work as a professional, the order matters: Block 1 first (tell it who you are), Block 2 second (let it see your work), Block 3 last (let it run without you). Most people skip straight to asking Claude questions and never set up the foundation — which is why they get search-engine results from a $20/month tool. The one-minute investment in Instructions pays back on every conversation for the rest of the year. You don't need a more powerful tool. You need to set up the one you already pay for.
If you want to go deeper after this, see our 5-level progression guide for using Claude in 2026, our complete guide to automating your small business with Claude Cowork, and our guide to putting Claude on autopilot with AI routines.
FAQ
Q: Do I need a paid Claude plan to set up these features? A: Instructions, Memory, Projects (up to 5), Connectors, and Skills are available on the free plan. Co-work and Scheduled Tasks require Claude Pro ($20/month) or above. The setup guide works regardless of your plan — start with Block 1 on a free account and upgrade when you're ready for Block 3.
Q: How long does it take to set up all 7 Claude settings? A: Block 1 (Instructions, Memory, Projects) takes about 6 minutes. Block 2 (Connectors, Skills) takes another 5–10 minutes depending on how many tools you connect. Block 3 (Co-work, Scheduled Tasks) takes 10–15 minutes to configure your first scheduled task. Total: under an hour for the full setup.
Q: What is the difference between Claude Instructions and Claude Memory? A: Instructions is what you explicitly write about yourself — it's a standing brief that never changes unless you edit it. Memory is what Claude learns on its own from your conversations and keeps forward. Both live in Settings, but Instructions is your input and Memory is Claude's learned context.
Q: Can Claude actually work on my files while I'm asleep? A: Yes — with Scheduled Tasks in Co-work, you point Claude at a folder and schedule a recurring job. It processes your files on schedule and prepares an update. This requires a Pro plan or above. Only 13% of knowledge workers have set up an automated workflow like this, per Section's 2026 study.
Q: Are Claude Connectors safe to use? A: Anthropic states that connector data is stored on Anthropic servers, protected by their security infrastructure, and they do not train models on your Gmail, Drive, or Calendar data. Each action Claude takes requires your explicit approval. However, Gmail is draft-only (Claude cannot send) and Microsoft 365 is read-only. Review the permission scope before connecting any tool. (Claude Help Center)
Q: How many Claude Projects should I create? A: Create one per client or per recurring area of work. A consultant might have one project per client; a marketer might have one per brand or campaign. Free accounts get 5; Pro accounts get unlimited. The point is separation — each room has its own context so conversations in one client's room don't bleed into another's.
Discussion
0 comments