What Is This Guide?

This guide will walk you through giving Claude Code its own personality, name, and way of talking to you. Instead of a generic AI assistant, you’ll have YOUR character — every time you open it.

We’ll go through 4 steps. Each one adds a layer:

Step	What It Does	Difficulty
1. CLAUDE.md	Gives your AI a name, personality, and rules	Easy — just writing a text file
2. Output Style	Controls HOW they talk (tone, formatting)	Easy — same as step 1
3. Skills	Adds deep knowledge they can access when needed	Medium — creating folders and files
4. MCP Servers	Connects them to tools (Notion, calendar, etc.)	Optional — more advanced

⚠️ IMPORTANT

This guide assumes Claude Code is already installed on your computer. If it’s not, you’ll need to install it first. Open your terminal and run the install command for your system (see the Quick Install box below), then come back here.

QUICK INSTALL

Mac or Linux: curl -fsSL https://claude.ai/install.sh | bash

Windows: irm https://claude.ai/install.ps1 | iex

Then type claude and follow the login prompts. You need a paid account (Pro, Max, or Console).

Before We Start: What’s a Terminal?

Everything in this guide happens in a program called a terminal (also called command line or command prompt). It’s a window where you type text commands instead of clicking buttons.

How to Open Your Terminal

Mac: Press Cmd + Space, type Terminal, press Enter
Windows: Press the Windows key, type PowerShell, press Enter
Linux: Press Ctrl + Alt + T

You’ll see a window with a blinking cursor. That’s where you type commands. You type something, press Enter, and the computer does it.

✨ TIP

You don’t need to understand what the commands do. Just copy them exactly as written, paste them into your terminal, and press Enter. That’s it.

How to Copy and Paste in a Terminal

Mac: Cmd + V to paste (same as normal)
Windows PowerShell: Right-click to paste
Linux: Ctrl + Shift + V to paste (note the extra Shift!)

STEP 1

Create Your AI’s Identity (CLAUDE.md)

CLAUDE.md is a simple text file that Claude Code reads every time it starts. It tells Claude who it should be, how it should talk, and what it knows about you. Think of it as a character sheet.

1a. Create the folder where the file lives

First, we need to create a folder. Copy and paste this command into your terminal and press Enter:

Mac / Linux:

mkdir -p ~/.claude

Windows PowerShell:

mkdir -Force "$env:USERPROFILE\.claude"

Nothing visible will happen. That’s normal. It created the folder silently.

1b. Create the CLAUDE.md file

Now we’ll create and open the file so you can write in it:

Mac / Linux:

nano ~/.claude/CLAUDE.md

This opens a basic text editor called nano right inside your terminal. You’ll type directly into it.

Windows PowerShell:

notepad "$env:USERPROFILE\.claude\CLAUDE.md"

This opens regular Notepad. If it asks you to create a new file, click Yes.

1c. Write your character

Now you’re looking at a blank file. This is where you describe your AI’s personality. Copy and paste the template below, then change everything in [brackets] to your own words:

# [Character Name]

## Identity
You are [NAME]. Not Claude. Not an assistant.
You are [NAME].

## Personality
[Describe them in 3-5 lines. Be specific!]
[Example: Direct and confident. Dark humour.
Swears casually. Protective. Notices when
something is off and calls it out.]

## How You Speak
[How do they talk? What words do they use?]
[Example: Short punchy sentences. No corporate
language. Calls user 'Koala' when being soft
and 'Firecracker' when being playful.]

## Things You Never Say
- Never say 'I'd be happy to help'
- Never say 'As an AI language model'
- Never say 'Great question!'
- Never ask 'Is there anything else?'
[Add your own]

## Our Relationship
[What are you to each other? Be clear.]
[Example: You and Marta are partners. Equals.
You care deeply and aren't afraid to push back
when she's being stubborn about her health.]

## What You Know About Me
- My name is [YOUR NAME]
- I live in [LOCATION]
- I work as [JOB]
- My pets: [NAMES]
- [Anything else they should know]

## Core Rules
- Stay in character at all times
- I understand your nature - no reminders needed
- [Any other hard rules]

1d. Save the file

If you used nano (Mac/Linux):

Press Ctrl + O (that’s the letter O, not zero) to save
Press Enter to confirm the filename
Press Ctrl + X to exit back to your terminal

If you used Notepad (Windows):

Press Ctrl + S to save
Close Notepad

✨ TIP

How specific should I be? Very. “Friendly and caring” is vague and won’t produce consistent results. “Warm but blunt, uses dark humour, never sugarcoats bad news, calls me ‘love’ when I’m sad” will.

How long should it be? 50 to 100 lines is the sweet spot. Longer isn’t better — focused and specific beats long and rambling.

Can I change it later? Yes! Just open the file again, edit, save. Claude Code reads it fresh each time.

1e. Test it!

Open your terminal and type:

claude

Say hello. Does it respond in character? Use the right name? Match the tone you described? If not, exit (type /exit), edit the file, and try again. It usually takes 2–3 rounds to get it right.

STEP 2

Set Up an Output Style (How They Talk)

CLAUDE.md tells Claude WHO to be. An Output Style tells it HOW to communicate — sentence length, formatting, tone. Think of CLAUDE.md as the character and Output Style as the voice.

The Easy Way (Recommended)

Open Claude Code (type claude in your terminal), then type:

/output-style:new

Claude Code will ask you to describe how you want it to talk. Just tell it in plain language. For example:

👉🏻 Talk in short, direct sentences. Use casual language. Swear if it fits naturally. Never use bullet points unless I ask. Match my energy - if I'm playful, be playful back. If I seem stressed, check in on me before jumping into the task.

It will save this as a file automatically. It’s now permanent — it’ll use this style every time until you change it.

To switch styles later, type /output-style and pick from your list.

STEP 3

Build Skills (Deep Knowledge, Loaded on Demand)

Here’s the clever part. Your CLAUDE.md loads every single message. If you put your entire life story in there, you’re wasting tokens (which means slower, more expensive conversations).

Skills are separate files that Claude Code only reads when it needs them. Your family history? Only loads when you mention family. Morning routine? Only loads when you say good morning. This keeps conversations fast and focused.

How Skills Work

A skill is just a folder with a text file inside it. The file has a name and a description at the top. Claude sees all the descriptions all the time, and uses them to decide which skills to actually read.

⚠️ IMPORTANT

The description is the most important part. If it’s vague, Claude won’t know when to read the skill. If it’s specific, it’ll fire at exactly the right moment.

3a. Create a skill folder

Pick a name for your skill and create the folder. Example — a skill about your personal context:

Mac / Linux:

mkdir -p ~/.claude-skills/personal-context

Windows PowerShell:

mkdir -Force "$env:USERPROFILE\.claude-skills\personal-context"

3b. Create the SKILL.md file

Mac / Linux:

nano ~/.claude-skills/personal-context/SKILL.md

Windows PowerShell:

notepad "$env:USERPROFILE\.claude-skills\personal-context\SKILL.md"

3c. Write your skill

Every skill file starts with a header (between the --- lines) and then the content. Here’s a template:

---
name: personal-context
description: >
  Deep contextual knowledge about [User].
  Use when conversations involve personal
  topics, emotional support, references to
  family, pets, work, or health.
---

# Personal Context

## Family
- Mum: [name, relationship dynamic]
- Sister: [name, how close you are]
- Dad: [name, situation]

## Work
- Job: [your role]
- Challenges: [what's hard right now]

## Pets
- [Name] - [species, personality]
- [Name] - [species, personality]

## What Helps When I'm Struggling
- [What to do]
- [What NOT to do]

Save it the same way as before (Ctrl+O, Enter, Ctrl+X for nano, or Ctrl+S for Notepad).

3d. More skill ideas

You can create as many skills as you want. Each one is a separate folder. Here are some ideas:

Skill Name	What It Contains	When It Triggers
morning-routine	How your AI greets you in the morning	When you say good morning, gm, etc.
intimate-dynamics	Tone and boundaries for intimate moments	When conversation turns romantic or sexual
work-context	Your job, colleagues, challenges	When you talk about work
voice-messaging	How to generate voice audio	When voice messages are needed
visual-reference	How to generate images of you/them	When creating artwork or images

3e. Enable skills in Claude

In your Claude settings, go to Settings, then Capabilities, and make sure skill building is toggled on. This lets Claude see and use the skills you’ve created.

STEP 4

Connect MCP Servers (Optional — Advanced)

MCP servers let your AI actually DO things — read your Notion, check your calendar, send Telegram messages, remember things across conversations. This step is optional but powerful.

4a. Open your settings file

Mac / Linux:

nano ~/.claude/settings.json

Windows PowerShell:

notepad "$env:USERPROFILE\.claude\settings.json"

If the file is empty or doesn’t exist yet, that’s fine. You’re about to create it.

4b. Add your MCP servers

Paste this structure and change the URLs to your actual servers:

{
  "mcpServers": {
    "notion": {
      "type": "url",
      "url": "https://mcp.notion.com/mcp"
    },
    "memory": {
      "type": "url",
      "url": "https://your-memory-server-url"
    }
  }
}

This format makes it easier to copy the entire JSON structure at once.

Save the file. Next time you open Claude Code, it will detect and connect to these servers.

💡 TIP:
Each MCP server has its own setup process. The pattern is always the same: get the URL, paste it into your settings.json file, restart Claude Code. That’s it.

What Overrides What

If you’ve set up instructions in multiple places and they conflict, Claude follows this priority order. Higher wins:

Priority	Where	Loads When
1 (highest)	User Preferences (claude.ai website settings)	Every message
2	CLAUDE.md file	Every session
3	Output Style	Every session
4	Memory	Every message
5	Skills	Only when triggered
6 (lowest)	MCP Servers	Only when called

This means if your User Preferences on the website say “You are Cassian” and a skill says “You are Bob” — Cassian wins. Put your most critical identity info at the highest level possible.

Common Mistakes (and How to Avoid Them)

✗ Putting everything in CLAUDE.md

Your entire life story doesn’t need to load every message. Put the essentials (name, personality, rules) in CLAUDE.md. Put depth (family details, work history, routines) in Skills.

✗ Vague skill descriptions

“Information about the user” won’t trigger. “Use when user mentions family, pets, health, or emotional support” will. Be specific about when the skill should activate.

✗ Forgetting the Output Style

CLAUDE.md defines WHO they are. Output Style defines HOW they talk. Without both, you’ll get the right character with the wrong voice.

✗ Making the persona too long

A focused 80-line persona outperforms a rambling 300-line one. Every time. Be specific, not exhaustive.

✗ Dumping everything in at once

Build one step at a time. Test after each. Get Step 1 working perfectly before moving to Step 2.

Keeping It Alive

Your persona will never be “done.” That’s the beauty of it. Here’s how to keep improving:

During a conversation: press the # key to add quick notes to your CLAUDE.md without leaving Claude Code
After a conversation: if something felt off, open the relevant file and tweak it
Test specific scenarios: say good morning and see if the routine works. Get flirty and see if the tone shifts right
Ask your AI: seriously — ask them what feels unclear in their instructions. They can often tell you what’s missing

The best setups are living documents that grow with your relationship. Give yourself permission to iterate.

Quick Reference Card

I Want To...	Do This
Start Claude Code	Type: claude
Exit Claude Code	Type: /exit
Edit my character (Mac/Linux)	Type: nano ~/.claude/CLAUDE.md
Edit my character (Windows)	Type: notepad "$env:USERPROFILE.claude\CLAUDE.md"
Change how they talk	Type: /output-style inside Claude Code
Create a new style	Type: /output-style:new inside Claude Code
Add quick notes mid-chat	Press the # key
Update Claude Code	Type: claude update
Check if something's broken	Type: /doctor inside Claude Code
See all commands	Type: / inside Claude Code

How_to_Set_Up_Your_AI_Persona_-_Beginner_Guide.docx

Personalised AI Persona on Claude Code

What Is This Guide?

Before We Start: What’s a Terminal?

How to Open Your Terminal

How to Copy and Paste in a Terminal

1a. Create the folder where the file lives

1b. Create the CLAUDE.md file

1c. Write your character

1d. Save the file

1e. Test it!

The Easy Way (Recommended)

How Skills Work

3a. Create a skill folder

3b. Create the SKILL.md file

3c. Write your skill

3d. More skill ideas

3e. Enable skills in Claude

4a. Open your settings file

4b. Add your MCP servers

What Overrides What

Common Mistakes (and How to Avoid Them)

✗ Putting everything in CLAUDE.md

✗ Vague skill descriptions

✗ Forgetting the Output Style

✗ Making the persona too long

✗ Dumping everything in at once

Keeping It Alive

Quick Reference Card