Stop guessing. Start configuring. This is the folder setup that turns Claude Code from a smart autocomplete into your personal marketing team.
What is Claude Code?
Claude Code is Anthropic’s CLI tool that lets Claude work directly in your terminal — reading files, writing content, running commands, and managing your entire workflow. But without the right configuration files, it’s flying blind.
These 5 folders and files tell Claude exactly how you work — your brand voice, your tools, your rules.
How to Get Started
- Install Claude Code if you haven’t already:
Terminal
$ npm install -g @anthropic-ai/claude-code
- Navigate to your project root:
Terminal
$ cd your-project
- Create the
.claudedirectory:
Terminal
$ mkdir -p .claude/commands
- Create your first config file:
Terminal
$ touch CLAUDE.md
That’s it. Now let’s fill these folders with the config that actually matters.
The 5 Folders
CLAUDE.md
Location: project root
Your project’s instruction manual that Claude reads before every session. This is where you define your brand voice, marketing rules, tools you use, and things Claude should never do.
Without this file, Claude makes assumptions. It might use the wrong tone, ignore your CTA style, or structure content differently than your brand expects. CLAUDE.md eliminates all of that by putting your rules in writing — once.
The result: Claude follows your exact brand standards from the very first prompt.
- Setting brand voice rules (tone, style, what to avoid)
- Defining your tech stack and preferred tools
- Specifying content format and structure
- Listing things Claude should never write or do
CLAUDE.md
# Project Instructions ## Brand Voice - Confident but not arrogant - No jargon, write for business owners - Always end with a CTA - Never use passive voice ## Tools We Use - HubSpot for CRM - Meta Ads + Google Ads - Canva for creatives - Notion for content calendar ## Content Rules - Headings: sentence case only - Paragraphs: max 3 lines - Always include keyword in H1
settings.json
Location: .claude/settings.json
Controls Claude’s permissions — what tools it can use, what it can’t touch, and environment-level config. Think of it as the security settings for your AI assistant.
By default, Claude asks permission for everything. That gets old fast. With settings.json, you whitelist the tools you trust and block the ones you don’t. You can also set environment variables like your preferred model.
The result: Fewer interruptions, tighter security, and a Claude that works within your guardrails.
- Allowing specific tools without confirmation prompts
- Blocking dangerous or irrelevant operations
- Setting environment variables for Claude’s behavior
- Configuring MCP server connections
settings.json
{
"permissions": {
"allow": ["Read", "Write", "Edit", "Bash", "Glob", "Grep"],
"deny": ["mcp__dangerous"]
},
"env": {
"ANTHROPIC_MODEL": "claude-sonnet-4-6",
"CLAUDE_CODE_MAX_TURNS": "100"
}
}
commands/
Location: .claude/commands/*.md
Custom slash commands that turn multi-line prompts into one-word shortcuts. Each .md file in commands/ becomes a /command you can run instantly — and you can pass arguments with $ARGUMENTS.
If you find yourself typing the same content brief, campaign checklist, or SEO instruction over and over — you’re wasting time. Turn any workflow into a repeatable slash command.
The result: Complex marketing workflows reduced to a single slash command.
- Blog post briefs you create for every article
- Instagram caption checklists you run on every post
- SEO audit templates you apply repeatedly
- Ad copy frameworks your team reuses
.claude/commands/write-blog.md
--- description: Write an SEO-optimised blog post allowed-tools: Read, Write --- Write a blog post on this topic: $ARGUMENTS Follow these rules: - H1 includes the primary keyword - Paragraphs max 3 lines - Include a CTA at the end - Pranav Veerani tone: direct, helpful, no fluff - Word count: 800–1200 words Output as clean HTML ready for WordPress.
Save this, then run /write-blog 5 Claude Code tips for marketers
hooks
Location: inside settings.json
Shell scripts that run automatically before or after Claude uses a tool. Every time Claude edits a file, writes content, or finishes a task — hooks trigger in the background without you thinking about it.
Hooks remove the manual steps you always forget. Auto-format content, run quality checks, send yourself a notification when a long task finishes — all without lifting a finger.
The result: Auto-QA, auto-formatting, and notifications on every Claude action.
- Auto-running spell/grammar checks after every content write
- Validating SEO rules before every file edit
- Sending yourself a Slack ping when Claude finishes a task
- Pushing content to your CMS automatically after creation
settings.json — hooks section
{
"hooks": {
"PreToolUse": [{
"matcher": "Edit|Write",
"command": "npm run lint -- --fix --quiet"
}],
"PostToolUse": [{
"matcher": "Write",
"command": "npx prettier --write $FILE_PATH"
}],
"Notification": [{
"command": "terminal-notifier -title Claude -message $EVENT"
}]
}
}
MCP Servers
Location: inside settings.json
Connects Claude to external tools — CRMs, analytics platforms, search engines, APIs, and anything else your marketing stack needs. MCP (Model Context Protocol) extends Claude’s reach beyond local files.
Out of the box, Claude can only work with your local files. Connect HubSpot and Claude can update your CRM. Connect Google Search Console and it can check your rankings. Connect any platform that has an MCP server.
The result: Claude goes from a content tool to a full-stack marketing partner.
- Querying your HubSpot or CRM data from Claude
- Searching the web for competitor research or trends
- Managing GitHub issues for your dev projects
- Connecting any API with an MCP server
settings.json — mcpServers section
{
"mcpServers": {
"supabase": {
"command": "npx -y @supabase/mcp-server",
"args": ["--project-ref", "your-project-id"]
},
"brave-search": {
"command": "npx -y @anthropic/brave-mcp",
"env": { "BRAVE_API_KEY": "your-key" }
},
"github": {
"command": "npx -y @anthropic/github-mcp",
"env": { "GITHUB_TOKEN": "ghp_xxx" }
}
}
}
Quick Reference
| Folder / File | Location | What It Controls |
|---|---|---|
| CLAUDE.md | Project root | Brand voice, rules, architecture |
| settings.json | .claude/settings.json | Permissions, env vars, tool access |
| commands/ | .claude/commands/*.md | Custom slash commands |
| hooks | Inside settings.json | Auto-run scripts on tool use |
| MCP Servers | Inside settings.json | External API and database connections |