All guides
ENGLISH9pMay 2026FREE TO READ

Getting Started with Claude Cowork

Set up the workspace, build your context files, and configure Cowork so it knows who you are and how you work. Five steps, with the prompts that build each file.

Read the guide

WHAT'S INSIDE

  • The folder structure that works, and why that one
  • The prompts that write your ABOUT ME files for you
  • CLAUDE.md: the standing context read before every session

Want the PDF?

The whole guide is open to read on this page. If you would rather keep a file to save, print or pass on, leave your email and it arrives with the newsletter.

You are signing up for Less Tools, More Skills. One email, roughly weekly. Unsubscribe from any issue.

Set up your workspace, build your context files, and configure Cowork to actually know who you are and how you work.

Access Claude Cowork

1 Download the desktop app

  • Go to claude.com/download and download for your computer
  • Cowork is only available in the desktop app - not in the browser, not on mobile

2 Make sure you have a paid account

  • Cowork requires a paid plan - Pro, Max, Team, or Enterprise
  • Starting at $20/month - upgrade at claude.ai → Settings → Upgrade
  • Free accounts will not see the Cowork tab

3 Open the app and switch to Cowork mode

  • Open Claude Desktop and sign in with your paid account
  • At the top of the app, look for the tabs: Chat | Cowork | Code
  • Click Cowork - you are now in task mode

4 Select a folder from your computer

  • Cowork will ask you to select a workspace folder
  • This is the folder Claude reads from and saves all its work to
  • How to set this up properly - right on the next page
⚠ Keep this in mind: The Claude Desktop app must stay open the entire time Claude is working. If you close it or your computer goes to sleep, the task stops. This is the most common reason tasks fail for new users.

Your Cowork Folder

Claude Cowork is all about how you set up your folder inside your computer. This is what separates a generic AI session from one that actually knows who you are, how you work, and what you are building.

Create this folder structure

  • Create a new folder on your computer named Claude Cowork
  • Inside it, create 4 subfolders exactly as shown below
📁 Claude Cowork/
├── 📁 ABOUT ME/        ← Most important. Claude reads this before every task.
│   ├── about-me.md
│   ├── anti-ai-writing-style.md
│   └── my-company.md
├── 📁 OUTPUTS/         ← Where Claude saves all its work
├── 📁 TEMPLATES/       ← Reusable formats Claude builds over time
└── 📁 PRESENTATIONS/   ← Optional, for slide-related work

Why the ABOUT ME folder matters

  • These are the only files Cowork reads automatically before every single task
  • The other folders are only read when you explicitly point Claude to them - saving your token budget
  • If these 3 files stay under 6,000 tokens total, Claude reads them completely - every session
  • If they grow too large, Claude summarizes loosely instead of reading carefully. Keep them lean.
The point: You never re-explain who you are or what you are working on. Claude already knows - because it read your files before starting. The context window is reserved for your actual task, not your profile.

File #1 - about-me

Who you are. How you think. How you want Claude to write for you. This is the highest-signal file in your setup - get it right and every Cowork session starts in the right place.

  • Keep it under 2,000 words - extract the patterns, throw away the raw transcripts
  • Every sentence should carry information. If it can be cut without losing signal, cut it.
Pro Tip: Dictate your answers instead of typing. Use Wispr Flow (free) - significantly faster and produces more natural, honest responses.

Don't have a file yet? Use this prompt

"You are building my about-me.md for my Cowork folder. It will be read by Claude before every session. Concise and
high-signal - under 2,000 tokens.
Your job: interview me with AskUserQuestion (15–20 questions), then compile into about-me.md.
Interview rules: one question at a time. If I'm vague, push back and ask for a specific example. Follow interesting threads
before moving on.
What to cover:
WHO I AM - What I do, my role and industry / who I work with or for / what a good week looks like
HOW I WORK - tools I use daily / how I start a task / my review process / what "done" looks like
WHAT GOOD LOOKS LIKE - best output I've produced recently and why / what separates great from average / what I notice
in others' work
WHAT I HATE - bad work examples and what specifically is wrong / patterns that make me cringe / what's usually off when
Claude writes for me
MY RULES - hard lines I won't cross / 2–3 non-negotiables every piece of my work must have
MY OPINIONS - what I believe that peers would push back on / what's overrated vs. underrated in my field
Output: Compile into markdown. No raw Q&A. Extract patterns, write as condensed prose and bullets.
Sections: Who I am / How I work / What good looks like / What I hate / My rules / Instructions for Claude.
Save as about-me.md in my ABOUT ME/ folder."

File #2 - anti-ai-writing-style

Your taste, tone, and rules - written as constraints for Claude. The words you hate. The sentence patterns that make you cringe. The formatting rules you care about.

  • Ban specific "AI words": delve, harness, tapestry, unleash, revolutionize - and any others that bother you
  • Set formatting rules: max paragraph length, no bullet overload, header preferences
  • Give Claude samples of your actual writing - LinkedIn posts, emails, reports
  • Be specific. "Write clearly" means nothing. "Never use em-dashes" is actionable.

File #3 - my-company

Your targets. Your strategy. What you are focused on. What you are saying no to. This file tells Claude what you are working toward right now - so it makes better decisions on every task.

Update this file when something actually changes - maybe once a quarter. Cowork does not need to know your Tuesday deadline. It needs to know your work.

Paste this in the same Cowork session as about-me (keep prompting):

"You are building my my-company.md for my Cowork folder. ONLY goals, strategy, and decisions - not who I am (that's
in about-me.md). Under 1,000 tokens.
Your job: interview me with AskUserQuestion (6–8 questions), then compile into my-company.md.
What to cover:
GOALS - top 2–3 goals this year with specific numbers or milestones / most important platforms or markets / the one metric
that would make this year a success / revenue or audience targets
DECISIONS - what I'm actively saying no to right now / what I recently stopped doing and why / where I'm spending most
time and energy this quarter / what I'm betting on that most peers aren't
Output: Short bullet points. No filler. No identity info.
Save as my-company.md in my ABOUT ME/ folder."

The OUTPUTS Folder

This is where Cowork works and saves all its deliverables.

  • One subfolder per project - Cowork organizes everything itself, automatically
  • Claude never reads from this folder on its own - that would consume your token budget unnecessarily
  • To reference a past deliverable: "Read the report in OUTPUTS/project-name"
  • Keep one subfolder per client, project, or campaign - you will thank yourself in three months

The TEMPLATES Folder

Cowork fills this folder over time. You do not set it up upfront - you build it as you work.

  • When Cowork builds something you like - a report, an email, a brief - say at the end of that session:
  • "Save this as a template in TEMPLATES/"
  • Next time you need something similar, Claude has a model to follow - no re-explaining, no recreating from scratch
  • Over time, your TEMPLATES folder becomes the memory of your best work

Templates + Global Instructions

  • The TEMPLATES folder does not fill itself - Claude only saves there when you explicitly ask it to
  • Your Global Instructions (next page) can tell Claude: "When I ask for a template to be saved, store it in TEMPLATES/"
  • This creates a compounding system: every session improves the next one
The pattern: ABOUT ME tells Claude who you are. OUTPUTS is where it works. TEMPLATES hold what worked. Global Instructions wire them all together.

Global Instructions

Your folder is not as good without Global Instructions. This is a prompt Claude Cowork reads before any task. Use it to explain how your folders work - because Claude does not know what your files mean or when to read them unless you tell it explicitly.

  • Go to Settings → Cowork → Edit Global Instructions
  • Paste the template below, then update the file descriptions to match your own
I usually start my Cowork session by pointing you to my Cowork folder.
Before any and every single task, you must read every file in ABOUT ME/:
- about-me: it's me, who I am, what I love and hate
- anti-ai-writing-style: I hate how Claude writes by default - always audit your output against this file before delivering
- my-company: where I work, my role, my current goals
Never read the folders OUTPUTS/ or TEMPLATES/ unless I specifically point you to a file.
Save all deliverables in OUTPUTS/ under a subfolder named after the project.
If the brief is unclear, use AskUserQuestion. Don't fill gaps with filler. Don't over-explain. Deliver the work.
Why this matters: Cowork reads your ABOUT ME files before every single task. If those 3 files are under 6,000 tokens total, Claude reads them completely - every session. You never re-explain who you are or what you are working on. The context window is for your actual task, not your profile. If your files grow too large, Claude starts summarizing instead of reading carefully - keep them lean and update them deliberately.

Ready to go further?

The best way to get better with Claude is to use it on real work - not practice prompts. Here are three things to explore once your folder is set up.

1 Skills

  • Saved instructions that tell Claude exactly how to handle a specific task - your voice, your format, your workflow. Use built-in skills or create your own for tasks you repeat often.
  • To explore: click Customize in the left sidebar → Skills

Examples: price-quotes · workshop-builder · morning-digest · shani-brand-voice

2 Plugins

  • Bundles of skills designed around a specific role - marketing, legal, sales, finance. Installing one gives you a whole set of relevant skills at once, pre-configured to work together.
  • To explore: click Customize → Browse plugins → pick a category

Examples: Marketing (campaign planning, SEO, content) · Sales (outreach, pipeline, call prep) · Brand Voice (guidelines, tone enforcement)

3 Connectors

  • Links Claude to your actual tools - email, Drive, Notion, Calendar. Without connectors, Claude works with what you tell it. With connectors, it works with what you already have.
  • To connect: click Customize → Connectors → choose a tool and authorize

Examples: Gmail · Google Drive · Notion · Google Calendar

AI Community Hub - Thousands of professionals are figuring this out together. Ask questions, share what's working, and see how others in real workplaces are using AI. facebook.com/groups/aicommunityhub

The CLAUDE.md File

CLAUDE.md is a file you place at the root of your Cowork workspace folder. Claude reads it before every session — before you type a single word. It is your standing context: who you are, what you are working on, how you work, and what rules Claude must follow. Define it once. Claude remembers every session.

Where to place it

  • At the root of your Claude Cowork folder — the folder you selected when setting up your workspace
  • Not inside a subfolder. Claude looks for it at the top level.
  • File name must be exactly: CLAUDE.md

Paste this into Claude Cowork to build your CLAUDE.md:

"Create a CLAUDE.md file for my workspace folder. Interview me to gather the information you need — one question at a time.

Cover: what I do and who I work with / my goals and priorities right now / how I like to work / my communication style and tone / rules you must always follow / which files in this folder to read before each task.

After the interview, write the file and save it to the root of this folder."

What to include — 10 key sections

  1. Project overview
  2. How I work
  3. Voice and tone
  4. Goals and priorities
  5. My audience / clients
  6. Active projects
  7. Folder and file structure
  8. Rules and hard limits
  9. Output preferences
  10. What good looks like

Project overview — the highest-value section.

Keep it to 2–3 short paragraphs. Claude needs a crisp mental model, not a brand story.
Bad: long origin story, "we value innovation," marketing copy with no practical relevance.
Good: "I am a GenAI consultant. My clients are CMOs at Israeli tech companies. Primary goal:
help teams move from AI curiosity to actual use."
Pro tip: Keep CLAUDE.md under 3,000 words. The context window is for your actual task — not your profile. Every sentence must carry signal. If it can be cut, cut it.

Keep this guide

The PDF arrives by email, along with Less Tools, More Skills: one practical GenAI skill a week.

Get the PDF