Resources / Guide

How to set up a Claude master workspace (free template)

By Taku Mambwere, Co-founder, Velorum. Updated 21 November 2026.

A Claude master workspace is one folder that holds everything you work on, with a subfolder for each area of your life and a CLAUDE.md file in each one. Claude reads CLAUDE.md files on its own, so it knows who you are, what you run and how you like things done without being told again (the root file loads every session, an area's file once Claude works in that folder). The free master document sets it up with you: attach it in Claude Code or Cowork, say interview me, and Claude builds the folders.

Get the free template
PartWhere it livesWhat it does
Root CLAUDE.mdWorkspace/CLAUDE.mdThe map: who you are, which folder is for what, house rules, read order
WORKING-WITH-ME.mdWorkspace/How you like Claude to work, your default tone, what it must never do
Area CLAUDE.md<Area>/CLAUDE.mdWhat the area is, who it serves, how it should sound, recurring jobs
Context files<Area>/context/about, me-here, strategy and current: the facts Claude needs for that area
Memorymemory/One note per session, a dated lessons file, dream reports
Skills.claude/skills/Jobs written down once: wrap-up, dream, new-skill
Archive_archive/Old files are moved here, nothing is deleted

Why a workspace beats a better prompt

Without a workspace, every new chat starts with you explaining yourself again: what you do, who the client is, how the email should sound. A workspace writes that down once, in files Claude reads on its own. It is not prompt engineering, it is context engineering: the more you work in it, the better Claude knows your work.

It works because of how Claude Code loads instructions. CLAUDE.md files in the folder where you start Claude, and in the folders above it, load at launch, and files in subfolders load once Claude works there. So the root file is always read, and each area's file is read when you work in that area.

Step 1: Make one folder, with a subfolder per area

Split your working life into areas, usually three to six: your job, each business or side project, content, studies, and a personal area for priorities and life admin. Each area gets its own folder.

Inside every area folder:

  • CLAUDE.md, the area's own instructions.
  • context/ with about.md, me-here.md, strategy.md and current.md.
  • plans/, scripts/, outputs/ and reference/, so files have a home.

Step 2: Write a CLAUDE.md for every folder

The root CLAUDE.md is the map. It says who you are, which folder is for what, the house rules (secrets never in files, nothing deleted) and the read order for every session.

Each area's CLAUDE.md says what the area is, your role, what matters in the next three months, how it should sound and what is sensitive. Keep each one short: Anthropic's advice is to target under 200 lines per CLAUDE.md, because longer files use more context and are followed less well.

Add a line whenever you correct Claude for the second time. That is the moment a preference becomes a rule.

Step 3: Let Claude interview you

You do not have to write any of this yourself. The free master document is a set of instructions for Claude. Attach it in Claude Code or Cowork and say interview me.

Claude first reads what it already knows about you (the folder, your existing instructions and memory, your connected tools) and says it back. Then it asks only what is missing, one question at a time, in four short rounds: you, your areas, each area, and what you are tired of explaining. It shows you the folder plan and builds only after you say yes.

Step 4: Add the rituals that make it learn

A workspace gets better only if each session leaves something behind. The download comes with three skills that do this. Skills live in .claude/skills/<name>/SKILL.md, and the folder name becomes the command you type.

  • wrap-up: at the end of a task, Claude writes a short session note, files the outputs, archives stray files and updates the area's current.md.
  • dream: every 5 session notes, Claude reads them together, turns patterns into lessons and proposes changes to your CLAUDE.md files. You approve every change.
  • new-skill: the second time you do the same job by hand, Claude writes it down as a skill, so it is done the same way every time after.

Step 5: Add skills and connectors you will actually use

Two additions are worth it:

  • Superpowers, an open-source skills library on Anthropic's official plugin marketplace: brainstorming before building, writing a plan, careful step-by-step work. In Claude Code: /plugin install superpowers@claude-plugins-official.
  • Connectors for the tools you already use, from the Anthropic Directory, so Claude can read and draft where you already work (how connectors work).

Claude Code, Cowork or a normal chat

The workspace runs in Claude Code or Cowork, because both can read and write files in a folder. In Cowork, skills come from the ones turned on for your claude.ai account, so add the workspace skills there if they do not show up.

In a normal claude.ai chat, Claude can still run the interview and write every file for you to save, but you need Claude Code or Cowork to work inside the workspace afterwards.

Keep secrets and deletions out of it

Two house rules protect you. Passwords, API keys and tokens never go in any file Claude reads; keep them in your system keychain or a password manager, outside the workspace. And nothing is deleted: old files move to _archive/ with the date in front, so any version can be found again. The template also offers optional save points, a private version history of the folder that is never pushed anywhere.

Where Velorum fits

You do not need us for any of this. The template is free, and Claude does the setup with you.

Where we help is the next step for a business: turning the jobs your team repeats into agents and automations that run on their own, connected to the tools you already use. If that is where you are, book a call.

Questions

What is a CLAUDE.md file?

A plain text (markdown) file with instructions for Claude. The CLAUDE.md in the folder where you start Claude loads every session, and the ones in subfolders load once Claude works there, so anything you write there does not need to be repeated in chat.

How long should a CLAUDE.md be?

Short. Anthropic's advice is to target under 200 lines per file. Put area details in that area's own CLAUDE.md and context files instead of one long root file.

Does the master workspace work in Cowork?

Yes. Cowork works in a folder like Claude Code. Cowork loads the skills turned on for your claude.ai account, so add the wrap-up, dream and new-skill skills there if they do not show up.

Can I use it in a normal claude.ai chat?

You can run the interview there, and Claude will write every file for you to save. To work inside the workspace afterwards you need Claude Code or Cowork.

What is the difference between CLAUDE.md and auto memory?

Auto memory is notes Claude saves for itself, on by default in local Claude Code sessions. CLAUDE.md files are instructions you own and can read. In the template, the workspace files are the source of truth, and the dream moves anything useful from auto memory into them.

Is the template free?

Yes. The master document, the starter folder and the PDF guide download straight from the page, with no email.

Sources

  1. Claude Code docs: how Claude remembers your project (CLAUDE.md, loading, 200 lines, auto memory)
  2. Claude Code docs: skills (where they live, Cowork)
  3. Claude Code docs: connect Claude to your tools
  4. Superpowers on GitHub
  5. Anthropic Directory of connectors

Want the jobs your team repeats turned into systems that run on their own?

Book a call