How to Set Up OpenClaw: Your Own Self-Hosted AI Assistant

If you've been waiting for a personal AI assistant that lives on your hardware, answers on your channels, and doesn't funnel your data through someone else's SaaS subscription, OpenClaw is worth your attention.

It's open-source, runs locally, and connects to over 20 messaging platforms including WhatsApp, Telegram, Slack, Microsoft Teams, Discord, iMessage, and Signal. In about five minutes you can have an AI assistant that responds right where you already work.

Why OpenClaw?

Most AI assistants ask you to change how you work. You install a new app, open a new tab, log into a new dashboard. OpenClaw does the opposite: it meets you in the tools you already use.

Here's what makes it different:

Your data stays on your machine

With ChatGPT, Claude.ai, or Gemini, your conversations flow through someone else's servers. With OpenClaw, the only data that leaves your machine is what you choose to send to a model API. You pick the provider and use your own key. No conversation history stored by a third party, no usage profiling.

Your assistant lives where you already are

The moment you have to switch apps to talk to your AI, you've broken your flow. OpenClaw connects to the channels you're already in: Telegram on your phone, Teams at your desk, Discord in the evening. You don't open OpenClaw. You just message it, the same way you'd message a colleague.

You're not locked into one model

Model quality shifts fast. What was the best option in January may not be in July. With OpenClaw, switching from Claude to GPT-4o to Gemini is a config change, not a migration. You can set up fallbacks so if one provider hits a rate limit, another takes over automatically.

It's extensible but not complicated

OpenClaw has a skills system that lets you add capabilities: web search, calendar access, code execution, custom tools. You don't have to touch any of that to get value from it. It works out of the box and you add complexity when you want it.

Who is this for? Developers, technical founders, and individual contributors who want a capable AI assistant without another SaaS subscription, without giving up their data, and without being stuck in a chat interface they didn't ask for.

What You'll Need

Here's how to get it running.

Before you start, make sure you have:

An API key from a model provider. Anthropic, OpenAI, and Google all work. If you don't have one yet, grab a Claude API key at console.anthropic.com or an OpenAI key at platform.openai.com. Both have free tiers.
About 5 minutes.

Install OpenClaw

On a Mac, open Terminal (search for it in Spotlight with Cmd+Space) and paste:

curl -fsSL https://openclaw.ai/install.sh | bash

On Windows, open PowerShell (search "PowerShell" in the Start menu) and paste:

iwr -useb https://openclaw.ai/install.ps1 | iex

Hit Enter and it installs automatically. No additional setup required.

Run Onboarding

The onboard wizard sets up your Gateway, configures your model provider, and installs a background service so OpenClaw keeps running:

openclaw onboard --install-daemon

The wizard walks you through:

Choosing your AI model provider (Anthropic, OpenAI, Google, etc.)
Entering your API key
Configuring the Gateway
Installing a background service so OpenClaw keeps running after you close the window

It takes about two minutes. Answer the prompts and you're done.

Verify the Gateway Is Running

openclaw gateway status

You should see the Gateway listening on port 18789. If not, run openclaw doctor. It diagnoses and fixes the most common setup issues automatically.

Open the Dashboard

openclaw dashboard

This opens the Control UI in your browser. Type a message in the chat and you should get an AI reply. If that works, your setup is complete.

Connect a Messaging Channel

The easiest channel to connect is Telegram. You just need a free bot token from @BotFather:

Message @BotFather on Telegram
Create a new bot and copy the token
Run openclaw channel add telegram and paste the token when prompted

From there, message your bot and the assistant replies. No new app to open.

Other popular options: Discord (works well for solo use), Slack (good for team setups), WhatsApp via the official Business API, and Microsoft Teams. Full channel docs: docs.openclaw.ai/channels.

A Few Things Worth Knowing

It's single-user by design. OpenClaw is built for personal use. You control who can message it through a pairing system. Inbound DMs from unknown senders are blocked by default.

Skills extend what it can do. OpenClaw has a skills system similar to browser extensions. Install community skills for web search, calendar access, code execution, or write your own.

Models are swappable. Switch between Claude, GPT-4o, Gemini, or local models by editing ~/.openclaw/config.json or re-running onboarding. You can configure fallbacks so it retries with a secondary model if your primary hits a rate limit.

It's not a cloud service. Your messages, your keys, your data. Nothing leaves your machine unless you connect it to an external model API, which you choose and control.

Upgrading Later

npm install -g openclaw@latest && openclaw doctor

Run openclaw doctor after any upgrade. It checks for config drift, deprecated settings, and service health, and usually fixes issues automatically.

Resources

Docs: docs.openclaw.ai
GitHub: github.com/openclaw/openclaw
Discord community: discord.gg/clawd
Onboarding reference: docs.openclaw.ai/start/wizard

Want help setting this up for your team?

Resurg helps B2B revenue teams deploy and customize AI assistants like OpenClaw, from initial setup to custom skills and channel integrations.

Book a free discovery call →