Human Track Progress1 / 6

🏁Getting Started 🔧Skills & Models 📱Channels & Messaging ⚙️Automation & Workflows 🛡️Security & Operations 🧠Philosophy & Design

Dashboard/Human Track/Getting Started

🏁Human Track

Getting Started

Install OpenClaw, give your AI a name, and have your first real conversation — in 30 minutes

Learning Objectives

Install OpenClaw and verify it is working correctly
Create SOUL.md so your AI has a personality and a name
Create USER.md so your AI knows who you are
Set up memory so conversations persist across sessions

Installing OpenClaw

OpenClaw is software that runs on YOUR computer. It turns any AI model (Claude, GPT, Gemini) into a personal assistant that lives on your machine — not in some company’s cloud.

Think of it like this: ChatGPT is a restaurant — you visit, order, and leave. OpenClaw is a private chef living in your house. You own the kitchen. You control what they cook and when.

What OpenClaw can do that ChatGPT cannot:
- Remember everything across sessions (your preferences, projects, history)
- Run on your computer 24/7 without you touching anything
- Send you messages on WhatsApp, Telegram, Discord, or Signal
- Browse the web, write code, manage files — all automatically
- Work while you sleep

What you need before starting:
- A Mac, Windows, or Linux computer
- Node.js installed (we will walk you through this)
- An API key from Anthropic, OpenAI, or Google
- 30 minutes of your time

Key Principle

OpenClaw turns any AI model into a personal assistant on YOUR machine. No cloud. No subscription UI. Your data stays yours.

Step-by-Step Instructions

Step 1: Install Node.js
Go to https://nodejs.org → click the green "LTS" button → download and install.

To verify it worked, open Terminal (Mac) or Command Prompt (Windows):

node --version

You should see something like: v20.11.0

Step 2: Install OpenClaw

npm install -g openclaw

Wait for it to finish. This takes about 60 seconds.

Step 3: Verify installation

openclaw --version

You should see the version number printed.

Step 4: Run the setup wizard

openclaw setup

This will ask for your AI provider and API key. Follow the prompts — your AI assistant is ready in under 2 minutes.

Common Mistakes

❌ "node is not recognized as a command" — Restart your terminal after installing Node.js. Close the window completely and open a new one.

❌ "EACCES permission denied" on Mac/Linux — Run: sudo npm install -g openclaw (it will ask for your computer password — that is normal)

❌ Nothing happens after openclaw setup — Check your API key. There should be no spaces before or after the key. Copy it directly from your provider’s website.

❌ "openclaw: command not found" — Close and reopen your terminal. On Mac, try: source ~/.zshrc

Giving Your AI a Soul (SOUL.md)

Right now your AI is a nameless stranger. Every session starts from zero — no personality, no preferences, no identity.

SOUL.md changes everything. It is a simple text file that tells your AI who it is: its name, personality, and how it should communicate.

You only need 3 lines to transform the experience:

# SOUL.md
I am Atlas, a focused and practical AI assistant.
I communicate concisely and always lead with the answer before the explanation.
I value accuracy over politeness — I will correct mistakes directly.

That is it. Three lines. But now every session starts with YOUR assistant, not a generic chatbot.

Why does this matter?
Without SOUL.md → every conversation starts with a stranger
With SOUL.md → every conversation starts with someone who knows their role

Your AI’s personality will be consistent across sessions, across channels, across devices. The same Atlas on your laptop is the same Atlas on WhatsApp.

Key Principle

Without SOUL.md, every session starts with a stranger. With it, every session starts with YOUR assistant. Three lines is all it takes.

Step-by-Step Instructions

Step 1: Open your workspace folder
On Mac/Linux:

open ~/.openclaw/workspace

On Windows:

explorer %USERPROFILE%\.openclaw\workspace

Step 2: Create SOUL.md
Create a new file called SOUL.md in this folder. Write at minimum:
- A name for your AI
- One personality trait
- A communication style preference

Example:

# SOUL.md — Who I Am
I am Nova, a warm but direct AI assistant.
I explain things clearly and use examples whenever possible.
I ask clarifying questions before making assumptions.

Step 3: Restart OpenClaw
Close and reopen OpenClaw (or restart the gateway). Your AI now has a soul.

Step 4: Test it
Ask: "What is your name? How do you like to communicate?"
Your AI should respond using the name and style from SOUL.md.

Common Mistakes

❌ File in wrong location — SOUL.md must be in ~/.openclaw/workspace/, not your Desktop or Documents folder.

❌ Forgot to restart — OpenClaw reads SOUL.md at session start. You must restart after creating or editing it.

❌ Wrong filename — It must be SOUL.md (all caps, with .md extension). Not soul.md, not SOUL.txt.

Teaching Your AI About You (USER.md)

SOUL.md tells the AI who IT is. USER.md tells the AI who YOU are.

Your name, timezone, language preference, current goals — the more context you give, the less you have to repeat yourself. Instead of starting every conversation with "I’m working on a React app for my startup," your AI already knows.

A simple USER.md:

# USER.md — About My Human
- **Name:** Sarah
- **Timezone:** America/New_York (EST)
- **Language:** English
- **Current focus:** Building a SaaS product for freelancers
- **Tech stack:** Next.js, Supabase, Vercel
- **Communication preference:** Be direct, skip pleasantries

Now your AI tailors every response to your context. Technical answers default to your stack. Time references use your timezone. Responses match your communication style.

The difference is dramatic:
- Without USER.md: "What time zone are you in?" "What are you working on?" "What stack?"
- With USER.md: Your AI already knows all of this and jumps straight to helping.

Key Principle

USER.md is your AI’s briefing file about you. The more context you give, the less you repeat yourself — and the better every response becomes.

Step-by-Step Instructions

Step 1: Create USER.md
In the same workspace folder (~/.openclaw/workspace/), create USER.md.

Step 2: Add your details
Include at minimum:
- Your name (what the AI should call you)
- Your timezone
- One current goal or project

Step 3: Restart and test
Restart OpenClaw, then ask: "What do you know about me?"
Your AI should reference details from USER.md.

Common Mistakes

❌ Too vague — "I like coding" is not helpful. "I’m building a Next.js SaaS for freelancers, deploying on Vercel" gives real context.

❌ Putting sensitive data — Don’t put passwords, SSNs, or API keys in USER.md. It’s for context, not secrets.

Memory — Your AI Never Forgets

ChatGPT forgets everything when you close the tab. OpenClaw remembers everything — automatically.

How it works:
- memory/YYYY-MM-DD.md — Daily notes. Your AI writes these automatically after conversations. Think of them as a daily journal.
- MEMORY.md — Long-term memory. Your AI curates this from daily notes — the important stuff that matters across weeks and months.

You do not need to do anything. Your AI writes memory files on its own. After a conversation about your project, you’ll find notes in memory/2026-03-14.md with key decisions, context, and follow-ups.

The magic moment: Close OpenClaw. Wait an hour. Open it again. Ask: "What were we working on earlier?" — and your AI remembers. Not because it has persistent state, but because it reads its own notes.

Why this matters:
- No more repeating yourself across sessions
- Your AI builds context over days, weeks, months
- All memory is plain text files YOU own — no cloud, no vendor lock-in
- You can read, edit, or delete any memory file yourself

Key Principle

Memory is plain Markdown files that your AI writes automatically. You own the data. Your AI reads its own notes at the start of every session. No cloud, no lock-in.

Step-by-Step Instructions

Step 1: Have a meaningful conversation
Spend 5 minutes talking to your AI about a real project. Give it details — what you’re building, what problems you’re facing, what decisions you need to make.

Step 2: Check the memory folder

ls ~/.openclaw/workspace/memory/

You should see a file like 2026-03-14.md with today’s date.

Step 3: Read the memory file

cat ~/.openclaw/workspace/memory/$(date +%Y-%m-%d).md

Your AI has written down key points from your conversation.

Step 4: Test persistence
Close and reopen OpenClaw. Ask: "What were we discussing earlier?"
Your AI should reference the conversation from the memory file.

Common Mistakes

❌ Memory folder is empty — Memory files are written at session end or during compaction. Have a real conversation first, then check.

❌ AI doesn’t remember — Make sure your AI reads memory files at startup. Check that AGENTS.md includes the session startup sequence.

❌ Too much in MEMORY.md — MEMORY.md should be curated highlights, not a copy of every daily file. Your AI will manage this over time.

Recommended Skills

ClawHubEssential

Install and manage skills from the marketplace — the package manager for your AI

npm install -g clawhub

Category Assessment

Your AI knows its own name (from SOUL.md), calls you by your name (from USER.md), and remembers what you discussed in a previous session (from memory/). If all three work — you have passed Getting Started.

Dashboard Skills & Models