How to Set Up OpenClaw: Your Own Personal AI Assistant

What OpenClaw Actually Is

Most AI assistants live inside someone else's app, forget you the moment you close the tab, and only do what that company decided to allow. OpenClaw is different. It is a free, open-source personal AI assistant — created by developer Peter Steinberger in late 2025 and, within months, the fastest-growing open-source project in GitHub history (over 300,000 stars). Its mascot is a small space lobster named Molty, and its whole philosophy ("the lobster way" 🦞) is that *your* assistant should answer to *you*.

What makes it genuinely useful for independent living is where it lives: inside the messaging apps you already open every day. You do not learn a new interface. You text your assistant on Telegram, WhatsApp, Signal, iMessage, Slack, Discord (it supports more than twenty channels), and it texts back — with persistent memory (it remembers your preferences, your people, your routines) and tool access (it can look things up, set reminders, draft messages, and take real actions you allow).

For anyone who wants a patient, always-available helper — to answer questions, read and summarize things, keep track of appointments and medications, or simply be someone to ask — OpenClaw is a personal aide you fully own and control. This guide walks you through setting one up.

Before You Start: What You Need

• A computer or a small server that can stay on. OpenClaw runs on macOS, Linux, or Windows. It can live on your everyday computer, or on an always-on machine (an old laptop, a mini PC, or a low-cost VPS) so it is reachable around the clock.
• Node.js. Version 24 is recommended (22.19 or newer also works). Do not worry if you do not have it — the quick installer can add it for you.
• An AI model "brain." OpenClaw itself is just the assistant framework; it needs a model to think with. The simplest path is an API key from a provider like Anthropic (Claude), OpenAI, or Google — you paste it in once during setup. (Advanced users can point it at a *local* model instead, so nothing leaves the house — more on that in the privacy section.)
• The messaging app you want to chat through (Telegram and WhatsApp are the most popular starting points).
• About 15 minutes. The core install and first chat take roughly five; pairing a messaging channel and tidying settings take a little more.

> If you self-host on a VPS, the floor is modest — about 2 CPU cores, 2 GB of RAM, and 2 GB of storage. For a containerized (Docker) install, aim higher: ~4 vCPU, 8 GB RAM, 40 GB SSD.

Step 1 — Install OpenClaw

The fastest method is a single command that detects your operating system, installs Node if it is missing, installs OpenClaw, and launches the setup wizard automatically.

1. Open your terminal (on Windows, you can instead use the native OpenClaw Hub app or the PowerShell installer; WSL2 also works). 2. Run the official one-line installer from the OpenClaw documentation at docs.openclaw.ai/install. (Always copy install commands from the official docs, never from a random forum post — this is your assistant's home, so start it from a trusted source.) 3. Let it finish. It will download OpenClaw and everything it needs, then drop you straight into onboarding.

Prefer a different route? OpenClaw also installs via Docker, Nix, or npm — all covered in the docs. For a first-timer, the one-line installer is the gentlest.

Step 2 — Run the Onboarding Wizard

When onboarding opens, it walks you through three simple choices:

1. Pick your model provider — Anthropic, OpenAI, or Google (or a local model). If you are unsure, any of the big three is a fine start; you can change it later. 2. Paste your API key. This is the credential that lets your assistant think. Keep it private (see safety below) — treat it like a password. 3. Set up the Gateway — the part of OpenClaw that connects it to your messaging apps. The wizard configures this for you.

That is the whole quick-start. Within a few minutes you can send your first message and get a reply — your assistant is alive.

Step 3 — Pair Your Messaging App

This is the step that makes OpenClaw feel like a real companion rather than a program: connect it to the app you already text in.

1. In OpenClaw's settings, choose a channel — for example Telegram or WhatsApp. 2. Follow the pairing prompt. For Telegram this usually means creating a bot token and pasting it in; for WhatsApp it means linking a number. The docs give exact, current steps per channel. 3. Send a "hello" from your phone. When it answers, you are done — your assistant now lives in your pocket, reachable like any other contact.

From here on, you never need the terminal again for everyday use. You just message your assistant.

Step 4 — Give It Memory and Skills

A fresh assistant is capable but blank. Spend a few minutes teaching it who you are:

• Tell it your basics — your name, your time zone, who the important people in your life are, how you like to be spoken to. It will remember.
• Give it your routines — "remind me to take my morning medication at 8 a.m.," "every evening, tell me tomorrow's weather and calendar." Persistent memory means you only say it once.
• Turn on the skills you want. OpenClaw supports optional skills and plugins — web search, reminders, note-taking, and more. Enable the handful that match how you will actually use it; skip the rest to keep things simple.

The goal is not to configure everything on day one. It is to give it enough that the *next* conversation already feels like it knows you.

Step 5 — Keep It Running Day and Night

For an assistant to be there whenever you need it, it has to stay awake.

• If OpenClaw lives on your everyday computer, it runs whenever that computer is on — fine for daytime use.
• For a truly always-on helper, install it as a background service (a "daemon"), which the onboarding wizard can set up. On an always-on machine — an old laptop left plugged in, a mini PC, or a small VPS — it then keeps running through restarts, ready at 3 a.m. as readily as 3 p.m.

That around-the-clock availability is much of what makes it a genuine aide rather than an app you have to remember to launch.

Safety, Privacy, and Cost (Read This Part)

An assistant this capable deserves a little care — none of it hard.

• Guard your API key like a password. Anyone who has it can spend on your account. Never paste it into a chat, a screenshot, or a shared folder. If you think it leaked, revoke and regenerate it from the provider's dashboard.
• Know where your words go. With a cloud model (Anthropic/OpenAI/Google), your messages travel to that provider to be answered — normal, but worth knowing. If you want nothing to leave your home, run OpenClaw against a local model instead (see our companion guide on setting up a local AI assistant with Ollama). That keeps every conversation on your own hardware.
• Watch the meter. Cloud models charge per use. Casual personal use is typically inexpensive, but heavy, always-on automation can add up — OpenClaw's own creator famously ran a bill into the thousands stress-testing it at scale. Set a spending limit in your provider's dashboard so there are no surprises.
• Start it from the official source. Install only from docs.openclaw.ai and the official github.com/openclaw/openclaw repository. It is your assistant's foundation; begin it from trusted ground.
• Give actions gradually. OpenClaw can *do* things, not just answer. Grant it new abilities one at a time as your trust grows, rather than all at once.

The Bottom Line

OpenClaw turns a free, open-source project into something quietly profound: a personal AI assistant that lives in the messaging app you already use, remembers what matters to you, and answers to no one but you. Install it, pair it to your phone, teach it your routines, and keep it running — and you have a tireless, private helper on call day and night. For anyone building a more independent daily life, having that kind of aide, fully in your own hands, is a genuinely powerful place to start.