Your First Claude Code Workspace

A workspace is where Claude does its work for your business. Technically, it is any folder you open a Claude Code session in -- run claude in a directory and Claude has read/write access to everything inside it. But practically, a workspace is the organised environment you build so that Claude understands your work, follows your conventions, and behaves consistently without you having to explain yourself in every session.

The difference between a configured workspace and a bare install is significant. A bare install is like hiring a highly capable assistant who just walked in off the street -- they can do almost anything, but they know nothing about your business, your clients, your file structure, or how you want things done. A configured workspace gives that assistant a complete onboarding pack that they read at the start of every working day, automatically.

Step 1 -- Create your CLAUDE.md

CLAUDE.md is the most important file in any Claude Code setup. Placed at .claude/CLAUDE.md (or the project root), it loads automatically at every session start -- no need to re-explain the project each time.

# My Project -- CLAUDE.md

## Overview
A Next.js web application for [describe your project].

## Tech Stack
- Next.js 15, TypeScript, Tailwind CSS
- Firebase for auth and database

## Conventions
- UK English in all output
- Components in src/components/, pages in src/app/
- Never commit .env files

## Key Files
- src/app/layout.tsx -- root layout
- src/lib/firebase.ts -- Firebase initialisation

Step 2 -- Recommended folder structure

├── .claude/
│   ├── CLAUDE.md              ← master workspace instructions
│   ├── settings.json          ← Claude Code config and permissions
│   ├── rules/                 ← one file per behavioural concern
│   │   ├── communication.md   ← language, tone, UK English
│   │   └── blocked-commands.md ← commands requiring confirmation
│   ├── skills/                ← reusable slash-command skills
│   └── hooks/                 ← event-triggered shell scripts
└── [your project files]

Step 3 -- settings.json

Every Claude Code workspace has a settings.json file at .claude/settings.json. This is a configuration file that controls Claude's behaviour -- think of it as your preferences panel. It determines what Claude is allowed to do automatically, what requires your approval, and which automated actions (hooks) run in the background.

{
  "permissions": {
    "allow": [],
    "deny": []
  },
  "hooks": {}
}

You do not need to populate this manually at first -- Claude Code will prompt for permissions interactively during the session and you can approve or deny each one. Add explicit rules here as your usage stabilises.

Common things you might configure in settings.json over time:

• Allowed operations -- file types or folders Claude can edit without asking each time
• Denied operations -- actions Claude should never take, regardless of what you ask
• Hooks -- shell scripts that run automatically when certain events happen (for example, running a spell-check every time Claude writes a document, or backing up state at the end of every session)

Step 4 -- Generate CLAUDE.md with /init

If you are working in an existing codebase, run /init in a Claude Code session. Claude will analyse the project structure and generate a draft CLAUDE.md reflecting what it finds. Review and edit the output -- it rarely captures conventions correctly on the first pass, but it provides a useful starting scaffold.

Slash commands like /init are shortcuts you type in the Claude Code conversation to trigger specific actions. Instead of typing "please analyse this project and generate a CLAUDE.md file", you type /init and Claude runs the process automatically. Other examples you will encounter frequently: /status (show all active projects), /help (list available commands), and /memory (view what Claude has stored about your preferences). Skills -- reusable workflows you define -- are also triggered this way, with commands like /meeting-summary or /proposal.

Why workspace structure matters

The folder structure and configuration files in your workspace are not bureaucracy -- they are how you communicate your working context to Claude. A well-organised workspace means:

• Claude finds things quickly. When you ask it to work on a client project, it knows where to look because the file structure is consistent and the CLAUDE.md describes it.
• Claude behaves consistently. Rules files mean Claude applies the same standards every session -- UK English, no sending documents without checking, always backing up before changes -- without you repeating them.
• Claude improves over time. As you add rules, refine your CLAUDE.md, and build skills for recurring tasks, the workspace accumulates your working knowledge. New sessions start with full context rather than a blank slate.

Starting with AI Business OS instead

The steps above describe building a workspace from scratch. That is the right approach if you want to understand every component and tailor the structure to a specific codebase or project. But for general business and knowledge work -- writing, research, client management, content production, recurring workflows -- there is a faster starting point.

AI Business OS is a pre-built Claude Code workspace template designed for solo operators and small businesses. Rather than building your CLAUDE.md, rules library, and skills from scratch, you clone a repository that already contains them -- then adapt to your own context.

What AI Business OS includes

• Master CLAUDE.md -- workspace identity, role definition, file organisation conventions, and a full reference to all supporting files
• Rules library -- separate files for communication standards, autonomy levels, credential handling, commit discipline, context management, document extraction, and more -- each governing a specific aspect of Claude's behaviour
• Skills framework -- a library of pre-built skills for common business tasks: LinkedIn posts, YouTube SEO, client proposals, assignment grading, social media orchestration, and dozens more -- each invocable as a slash command
• State management -- a JSON-based project tracking system with session history, milestones, and cross-session continuity built in
• Session commands -- /day, /night, /sync, and /status for structured session start, end-of-day wrap-up, git synchronisation, and project health reporting
• MCP server configuration -- pre-configured connection patterns for calendar, email, Notion, Slack, accounting, and other common business tools
• PRIMA Memory -- an optional companion MCP server that records every session to a local database, enabling cross-session context retrieval and session resumption by name or date

Build from scratch or use AI Business OS?

Both approaches use the same Claude Code components. AI Business OS is not a different product -- it is a starting configuration that would take significant time to build from scratch, provided as a template. See GenAI Skills Academy for setup guides and access.