hello@abiemaxey.com
Open Claw
Most people use AI through someone else's interface. This playbook builds yours from scratch ~ self-hosted, multi-channel, with skills, memory, voice, and smart routing. The full architecture of an AI assistant that belongs to you.
~ The Problem
You are renting someone else's AI.
Every time you use ChatGPT or Claude through their app, you are on their terms. Their pricing, their rate limits, their data policies, their model choices. You have no control over the infrastructure.
OpenClaw flips that. Your server, your models, your rules. One gateway that connects to every chat app you use, running whichever AI handles the task best ~ and it costs less than most subscriptions.
$ openclaw onboard --deploy
› Initializing gateway...
✓ VPS provisioned (Hetzner ~ 4GB)
✓ OpenClaw installed (v3.2.1)
✓ Telegram channel connected
✓ Claude API key loaded
✓ Smart routing enabled (Claude → Gemini → Ollama)
✓ Memory workspace initialized
✓ Security allowlist active
⚠ Monthly estimate: ~$8/mo (VPS + API usage)
› Your AI assistant is live.
✓ Gateway ready. You own this.
~ What's Inside
The 4 phases
Ground Zero~ From nothing to a running instance.
You have never deployed a server. That is fine. This phase takes you from zero to a working OpenClaw instance with security locked down ~ no Docker experience required, just a terminal and 20 minutes.
What OpenClaw Actually Is
OpenClaw is an open-source gateway that connects your favourite chat apps ~ Telegram, WhatsApp, Discord ~ to AI models like Claude, GPT, Gemini, and local models via Ollama. It runs on your own server, not someone else's cloud. That means you own the data, control the costs, and decide which models handle which tasks. This lesson covers the architecture at a high level ~ gateway, channels, skills, memory, models. Not a pitch, a schematic. By the end you know exactly what you are building and why owning the infrastructure changes the economics of AI.
The Architecture
At its core, OpenClaw is a gateway. Messages come in from chat apps (channels), get processed through a routing layer that picks the best model for the task, and the response goes back out. In between, it has access to memory (persistent context across sessions), skills (installable capabilities from ClawHub), and a workspace that ties everything together.
Why Self-Host?
When you use ChatGPT or Claude through their apps, you are locked into their pricing, their rate limits, their data policies, and their model choices. With OpenClaw, you pick the model per task ~ Claude for deep thinking, Gemini for cheap chat, Ollama for free local inference. You control the data. You set the rules. And the monthly cost is often less than a single AI subscription.
What You Will Build
By the end of this playbook, you will have a live AI assistant running on your own server, connected to your phone via Telegram, with skills installed, memory enabled, smart routing configured, and a maintenance system that keeps it running long-term. Not a demo ~ a production system.
📹 What Is OpenClaw ~ Metics Media
Your First Deployment
Choosing a VPS provider, spinning up the server, running the install, and getting your credentials. This lesson walks through the full deployment from an empty server to a running OpenClaw instance. Hetzner, DigitalOcean, Vultr ~ each has trade-offs. The interactive tool below helps you estimate what you actually need based on your intended usage. No Docker experience required ~ just a terminal and 20 minutes.
Choosing a VPS Provider
You need a Linux server with at least 2GB RAM. Hetzner is the cheapest for European hosting (~€4/mo for a CX22). DigitalOcean and Vultr are solid alternatives with data centers worldwide. For most users running 1-2 channels with standard skills, a $5-6/mo VPS is plenty. The estimator below helps you right-size based on your actual needs.
The Installation
OpenClaw ships as a single install script. You SSH into your server, run the command, and it handles everything ~ pulling the runtime, setting up the database, generating your credentials. The whole process takes about 5 minutes if your VPS is already running.
# SSH into your VPS
ssh root@your-server-ip
# Run the OpenClaw installer
curl -fsSL https://get.openclaw.ai | sh
# The installer will output your credentials
# SAVE THESE ~ you will need them in the next step
After Installation
Once the install completes, you will see your admin credentials in the terminal output. Copy them somewhere safe immediately. The next lesson covers your first conversation and locking down security ~ but first, verify the install worked by checking the status.
# Check OpenClaw is running
openclaw status
# You should see:
# ✓ Gateway: running
# ✓ Database: connected
# ✓ API: listening on :3000
📹 Where to Run OpenClaw ~ VPS Setup
Message Volume
Channels
Voice Mode
Skills Count
First Conversation & Security
Your first message to OpenClaw, the bootstrap flow, and then immediately locking it down. Security is not a phase you get to later ~ it is the first thing you do after hello. This lesson covers the onboard command, credential rotation, allowlists, mention requirements, and the essentials that keep your instance from becoming someone else's. If it is on the internet, it needs to be locked down. Period.
The Bootstrap Flow
Your first interaction with OpenClaw is the onboard command. This is where you set your master password, configure the admin channel, and verify the system responds. It takes about 2 minutes and confirms everything from the previous lesson is wired up correctly.
# Start the onboard flow
openclaw onboard
# Follow the prompts to:
# 1. Set your admin password
# 2. Configure your first channel
# 3. Send a test message
Security Essentials
If your OpenClaw instance is on the internet, it needs to be locked down. Security is not a phase you get to later ~ it is step one after hello. The basics: rotate your default credentials, set up an allowlist so only your accounts can interact with the bot, enable mention-only mode in group chats so the bot does not respond to every message, and configure rate limits to prevent abuse.
# Rotate default credentials
openclaw config set admin.password <new-password>
# Enable allowlist ~ only these users can talk to the bot
openclaw config set security.allowlist ["your_telegram_id"]
# Mention-only mode for group chats
openclaw config set channels.telegram.mention_only true
What Happens If You Skip This
Your bot becomes an open proxy. Anyone who finds it can use your API keys, rack up charges, and access anything the bot has access to. This is not theoretical ~ bots with exposed endpoints get found within hours by automated scanners. Lock it down before you do anything else.
📹 Security Essentials ~ Locking Down Your Instance
Channels & Skills~ Connect it to the real world.
An AI that only lives in a terminal is a toy. This phase connects OpenClaw to real chat apps and installs the skills that turn it from a chatbot into an operator.
Connecting Your First Channel
Telegram is the fastest channel to wire up. This lesson walks through creating the bot via BotFather, connecting it to OpenClaw, and sending your first message through a real chat app. Once one channel works, the pattern repeats for WhatsApp, Discord, and anything else. The moment you message your AI from your phone and get a real answer back ~ that is when it clicks.
Why Telegram First
Telegram is the fastest channel to set up because BotFather makes bot creation instant ~ no approval process, no business verification, no waiting. You create a bot, get a token, paste it into OpenClaw, and you are live. The whole process takes under 5 minutes. Once one channel works, the pattern repeats for WhatsApp, Discord, and anything else.
Creating the Bot
Open Telegram, search for @BotFather, and start a conversation. The /newbot command walks you through naming your bot and gives you an API token. That token is what OpenClaw uses to send and receive messages through your bot.
# In Telegram, message @BotFather:
# /newbot
# Follow prompts to name your bot
# Copy the API token
# In your terminal:
openclaw channel add telegram
# Paste your bot token when prompted
# Test it ~ send a message to your bot in Telegram
The Moment It Clicks
When you message your bot from your phone and get a real answer back ~ powered by Claude or whichever model you configured ~ that is when it clicks. This is not someone else's chatbot. This is your infrastructure, on your server, responding through your channel. Every message after this is running through a system you own.
📹 Connecting Telegram ~ Full Walkthrough
Understanding & Installing Skills
Skills are what turn OpenClaw from a chatbot into an operator. ClawHub is the marketplace ~ think npm for AI agent capabilities. This lesson covers how to evaluate skills, install your first one with npx, and understand what is happening under the hood. The marketplace is growing fast ~ knowing how to read a skill listing is a permanent advantage. Use the tool below to build your starter stack based on what you actually need.
What Are Skills?
Skills are modular capabilities you install into OpenClaw. Think of them like apps for your AI ~ each one adds a specific ability. A web search skill lets your bot search the internet. A calendar skill lets it read and create events. A code review skill lets it analyze code snippets. They are installed via npx from ClawHub, the skills marketplace.
Evaluating a Skill
Not all skills are equal. Before installing, check the ClawHub listing for: how many installs it has, when it was last updated, what permissions it requires, and whether it has a clear README. A skill with 500+ installs and a recent update is a safer bet than one with 3 installs from six months ago. Read what data it accesses before you install it.
# Browse available skills
npx clawhub@latest search web-search
# Install a skill
npx clawhub@latest install web-search
# List installed skills
openclaw skills list
# Remove a skill
openclaw skills remove web-search
Building Your Starter Stack
You do not need 20 skills on day one. Start with 2-3 that match your actual daily use. The interactive tool below helps you pick based on what you will actually use your assistant for. You can always add more later ~ but a focused starter stack means less configuration overhead and faster time to value.
📹 Understanding & Finding Skills on ClawHub
What will you use OpenClaw for?
Memory, Identity & the Workspace
Memory is what separates a tool from an assistant. This lesson covers how OpenClaw stores context across sessions, how to shape its identity and persona so it responds the way you need, and how the workspace ties everything together. A well-configured workspace is the difference between an AI that forgets you every session and one that compounds knowledge over time. Identity is not cosmetic ~ it is functional.
How Memory Works
Without memory, every conversation starts from zero. OpenClaw's memory system stores context across sessions ~ what you talked about, decisions that were made, preferences you expressed. It uses a combination of short-term session context and long-term memory that persists between conversations. This is what makes the difference between a tool you use and an assistant that knows you.
Shaping Identity
Identity is not just a name and a tone. It defines how your assistant interprets ambiguous requests, what it prioritizes, and how it structures responses. A well-configured identity means less prompting and better outputs. Set the role, set the constraints, set the communication style ~ and let it compound over time.
# Set your assistant's identity
openclaw config set identity.name "Atlas"
openclaw config set identity.role "Personal research assistant"
openclaw config set identity.style "concise, direct, no fluff"
# Enable persistent memory
openclaw config set memory.enabled true
openclaw config set memory.max_entries 1000
# View stored memories
openclaw memory list
The Workspace
The workspace is where everything ties together ~ memory, skills, identity, and channel configuration. Think of it as the operating environment for your assistant. A clean workspace with clear identity, curated memory, and focused skills will outperform one with 30 skills and no direction. Quality of configuration beats quantity every time.
├── AGENTS.md ← Operating instructions (loaded every session)
├── SOUL.md ← Personality and values (loaded every session)
├── USER.md ← Info about YOU (loaded every session)
├── IDENTITY.md ← Bot's name, emoji, vibe
├── TOOLS.md ← Notes about available tools
├── HEARTBEAT.md ← Periodic check-in tasks
├── BOOT.md ← Startup checklist (optional)
├── MEMORY.md ← Long-term curated memory
├── memory/ ← Daily logs (today + yesterday loaded)
│ ├── 2026-02-26.md
│ └── ...
└── skills/ ← Workspace-specific skills
📹 Memory, Identity & the Workspace
Operations~ Make it run like production.
Your instance works. Now make it smart. This phase covers the operational layer ~ automation, model routing, cost optimization, and the maintenance habits that keep a self-hosted system running clean.
This phase is locked
Subscribe to unlock all 3 lessons in this phase plus full progress tracking.
07 ~ Heartbeat, Cron Jobs & Automation
15 min08 ~ Model Selection & Cost Optimization
20 min09 ~ Diagnosing Errors & Maintenance
15 minAdvanced~ Voice, sub-agents, and the long game.
Your instance is stable, secure, and cost-optimized. This phase unlocks the advanced capabilities ~ voice interaction, multi-agent architecture, and the operational mindset that treats your AI infrastructure as a long-term asset.
This phase is locked
Subscribe to unlock all 3 lessons in this phase plus full progress tracking.
10 ~ Voice Mode
15 min11 ~ Sub-Agents & Multi-Agent Architecture
20 min12 ~ Updates, Recovery & the Long Game
14 min~ What comes next
The next OS
The Talent Stack
Stop competing on price. Become irreplaceable.
You have the AI infrastructure. Now package the skills that make you indispensable ~ the operator clients renew, refer, and never want to lose.
Still Running
Build leverage, stack income, leave on your terms.
Still in the machine? This playbook maps the exit ~ from cage audit to clean resignation, with the income proof to back it up.
The builder's diaries.
No fluff. No BS. Just the systems, visas, and real-world operations of building a life across borders.
- Nomad visa breakdowns ~ no fluff
- Spain DNV updates as they happen
- Remote income systems that actually work
- Honest field notes from the road
Growing Community
Drop your email.
Unsubscribe any time. No spam, ever.