Build Your Personal AI Agent in 30 Days

The Master Guide

No Coding Required. Written by Tim Pittman.

You Don't Need to Be Technical. You Just Need to Be Willing. OpenClaw lets you build powerful automation without writing a single line of code.

// WHY THIS GUIDE?

Freedom from Routine.

You're busy. You don't have time to learn programming just to automate your work. But you also don't want to feel left behind as AI transforms everything.

Overwhelmed?

Ideal if manual tasks are eating your day and you want to work smarter, not harder.

Feeling Lost?

Designed for professionals who've tried AI tools but felt overwhelmed by the technical complexity.

// WHAT YOU'LL LEARN

1. Rapid Setup

Install and configure OpenClaw, navigate the dashboard, and connect messaging apps (Telegram, Slack, WhatsApp).

2. Model Deep Dive

Understand Claude, GPT, and Minimax. Choosing the right "brain" for the right cost and task.

3. Building a Soul

Learn why identity matters and how to edit your agent's personality files (SOUL.md).

4. Memory & Context

Set up memory that persists across sessions so your bot actually remembers you.

// SECTION 1

Setup

Installation

OpenClaw runs on Mac, Linux, or Raspberry Pi. Here's your quickstart command sequence:

                
                # Clone the repo
git clone https://github.com/openclaw/openclaw.git
cd openclaw

# Install dependencies
npm install

# Run the setup
npm run setup

Connect Your Messaging App

Message @BotFather on Telegram
Create a new bot with /newbot
Copy API token into your config

Slack

Go to Slack API → Create App
Enable Bot Token Scope
Install to workspace & copy Token

Supports WhatsApp Business API
Connect via QR code scanning
Follow in-app setup wizard

Dashboard Navigation

Agents

Your active AI personas.

Skills

Available bot capabilities.

Memory

Visual conversation history.

Settings

Global app configuration.

Your First Agent — Daily Briefing

Paste this prompt to start a professional morning routine:

                
                You are my Personal AI Assistant. Every morning at 8am, review my calendar, tasks, and any important updates. Then provide a brief, actionable summary covering:
- Top 3 priorities for today
- Any meetings that need preparation
- Follow-ups or replies I might be missing
Keep it short, clear, and actionable. Format as a bullet list.

Agents vs Sub-Agents: Agents are independent workers with personality. Sub-Agents are task-specific contractors spawned by your main agent for a single job.

// SECTION 2

Understanding Models

One of OpenClaw's superpowers is switching between AI models depending on what you need.

Anthropic (Claude)

Best for: Nuanced reasoning, creative writing, deep analysis.
Models: Haiku (fast), Sonnet (balanced), Opus (quality).

OpenAI (GPT)

Best for: Code, general speed, complex instructions.
Models: 4o Mini (cheap), 4o (balanced), o1 (reasoning).

Minimax

Best for: Speed and cost-sensitive tasks.
Models: MiniMax-M2.5 series.

When to Switch Models

Task	Recommended Model
Quick reminders, simple lookups	Mini / Haiku
Writing, editing, brainstorming	Sonnet / GPT-4o
Complex analysis, deep reasoning	Opus / o1

Cost Management Tips: Default to cheaper models for routine tasks. Set usage alerts in your configuration and review token usage weekly.

// SECTION 3

The Bot's Soul

Understand and customize your agent's identity. This is where most people get stuck.

SOUL.md

Defines WHO your agent is. Personality, values, and voice.

USER.md

Defines WHO YOU ARE. Your preferences, context, and goals.

TOOLS.md

Your LOCAL setup. Camera names, SSH aliases, preferences.

The File Structure

workspace/
├── SOUL.md          ← Identity
├── USER.md          ← Your Context
├── TOOLS.md         ← Environment
├── MEMORY.md        ← Long-term Facts
├── BRAIN.md         ← Active Projects
├── memory/          ← Daily Logs
└── agents/          ← Sub-agents

Key Principle: SOUL = WHO | SKILL = HOW | TOOLS = WHERE

How to View and Edit

View: Ask your agent "show me SOUL.md"
Edit: Say "update SOUL.md to include..." or edit files directly
New Agents: Add folders in agents/ with their own SOUL.md

// SECTION 4

Memory & Context

Without proper setup, your agent forgets everything between sessions. This is critical for high-fidelity work.

SOUL.md

Identity & Personality. Loaded every session to set the tone.

USER.md

Directly shapes responses based on your communication style. Most powerful behavioral tool.

MEMORY.md

Long-term curated facts. Aim for a one-page onboarding doc format.

Daily Logs (YYYY-MM-DD.md)

Running context. Prune periodically into MEMORY.md to keep context clean.

What Happens Automatically

Start: Reads SOUL, USER, MEMORY, and recent daily logs.
Compaction: Triggers a flush when approaching context limits.

⚠️ The Big Gotcha

If you end a session without explicitly saving, most decisions are gone. The model keeps things in RAM and only writes when told or when compaction fires.

The Four Levels of Saving

Level	Command	Persistence
Level 1	"Remember this"	Weak (Current session only)
Level 2	"Save to daily log"	Good (Recent context)
Level 3	"Update MEMORY.md"	Best (Long-term facts)
Level 4	"Add to USER.md"	Strongest (Behavioral rules)

The Golden Habit: Before closing, say: "Update the daily log and MEMORY.md with what we worked on today, decisions made, and what's pending."

Advanced Configuration

{
  "compaction": {
    "reserveTokensFloor": 20000,
    "memoryFlush": {
      "enabled": true,
      "softThresholdTokens": 4000,
      "prompt": "Write lasting notes to memory/YYYY-MM-DD.md"
    }
  }
}

// SECTION 5

Skills

Skills are reusable tool sets that give your agent new capabilities. Reusable and modular.

Weather

Forecast retrieval integration.

Healthcheck

Automated security auditing.

Things

Task management and calendar logic.

Creating Custom Skills

Create a SKILL.md file in your skill folder with this structure:

# My Custom Skill
## What This Does
[Purpose]
## How to Use
- Command 1
- Command 2
## Tools Used
- toolA

// ADVANCED PATTERNS

Task Type	Recommended Approach
Quick one-off lookup	Direct Execution
Complex multi-step project	Sub-Agent Deployment
Specialized recurring tool	Custom Skill Creation
Ongoing permanent role	Full Dedicated Agent

Expanding the Brain

Obsidian: For long-term knowledge and meeting notes.
Notion/Notes: Via built-in connectors or API skills.

Security First

Never expose API keys in prompts. Run locally so data stays on your machine. Review agent outputs before sending externally.

// FIN

You have reached the end of the Master Guide. Go forth and build.