OpenClaw: Self-Hosted AI Assistant You Can Message From Anywhere

What Is OpenClaw and Why Self-Host It?

OpenClaw is an open-source, self-hosted Gateway that connects chat apps (WhatsApp, Telegram, Discord, iMessage, etc.) to AI agents. You run one Gateway process on your own server and it becomes the bridge between your messaging channels and your AI agent runtime — so the “interface” is just messaging. (OpenClaw )

If you’ve ever wished your AI could actually do things — clear inboxes, send emails, manage calendars, or run multi-step workflows — OpenClaw is built for exactly that.

This article explains how OpenClaw works and walks you through a beginner-friendly deployment path: deploying OpenClaw on Zeabur, where you can rent a server and ship your stack in minutes.

Key Features of OpenClaw

From the official docs :

• Self-hosted: runs on your hardware, your rules
• Multi-channel: one Gateway can serve multiple chat platforms
• Agent-native: built for tool use, sessions, memory, and routing
• Control UI: browser dashboard for chat, config, and session management
• Open source (MIT licensed)

How OpenClaw Works (Architecture Overview)

OpenClaw revolves around the Gateway:

• It’s the single source of truth for sessions, routing, and channel connections. (Docs )
• The Control UI (browser dashboard) lets you chat, configure settings, and manage sessions.
• The Gateway WebSocket defaults to binding to loopback (localhost) on port 18789, and remote usage is typically done via SSH tunnels or tailnet/VPN patterns. (Remote Access Guide )

That default “loopback-first” posture is one reason people like self-hosting it: you can keep sensitive control surfaces private and only expose what you intend.

Why Deploy OpenClaw on Zeabur?

You can run OpenClaw locally or on any VPS, but Zeabur is a great fit when you want the convenience of a PaaS plus the option to rent and manage servers easily.

Key reasons:

1. Fast onboarding: create an account, create a project, deploy services. (Zeabur Get Started )
2. Flexible compute:
   • Shared Cluster: simplest “I don’t want to think about servers” option
   • Dedicated Server: a server that belongs only to you (better isolation + stability), and Zeabur supports purchasing one directly from the dashboard. (Zeabur Dedicated Server )
3. Production basics built-in: environment variables management, domains, service logs, etc. (Zeabur Variables )

The Easiest Path: Zeabur’s OpenClaw Template

Zeabur provides an official OpenClaw template  that’s already wired up to run without you manually executing the setup wizard. It uses an OpenClaw container image and includes usage instructions right in the template page.

Recommended Resources

The template suggests:

• Minimum: 2 vCPU / 4 GB RAM
• Recommended: 4 vCPU / 8 GB RAM

Step-by-Step: Deploy OpenClaw on Zeabur

1) Create a Zeabur Account and Project

Follow Zeabur’s standard flow:

• Create an account at zeabur.com 
• Create a project and pick a region (Get Started Guide )

2) Choose Your Compute: Shared Cluster vs Dedicated Server

• Shared Cluster is simplest: you just pick a region and deploy. (Dedicated Server Docs )
• Dedicated Server is appealing when you want more control and isolation; Zeabur supports buying a dedicated server from their dashboard.

If you’re planning to expose the Control UI to the internet, Dedicated Server can also make it easier to apply network restrictions (more on security below).

3) Deploy OpenClaw from the Template

In Zeabur Templates , pick OpenClaw, then deploy it.

What’s nice here:

• The template is pre-configured and ready to use — you don’t need to run the setup wizard unless you want advanced config.

4) Open the Web UI Using the Token Link

After deployment:

1. Copy the “Web UI (with token)” URL from the Instructions tab
2. Open it in your browser
3. First login requires the token; afterwards you can access normally

Treat this token like a password.

5) Configure Your AI Model Provider

The template supports two common paths:

Option A — Zeabur AI Hub:

If you enter the API key during deployment, the template defaults to a Zeabur-hosted model. You can also add ZEABUR_AI_HUB_API_KEY later in the Variables tab and restart the service.

Option B — Other Providers (Anthropic, OpenAI, etc.):

If you don’t set Zeabur AI Hub, configure your provider’s API key via the Web UI settings or environment variables. For example, OpenClaw supports ANTHROPIC_API_KEY for Anthropic models. (Anthropic Provider Docs )

6) Optional: Connect Telegram

A quick way to start chatting with your AI from anywhere:

1. Create a Telegram bot via @BotFather
2. Add TELEGRAM_BOT_TOKEN in Zeabur environment variables
3. Restart the service
4. Send /start to your bot — you’ll get a pairing code
5. Approve pairing via the Web UI chat or the container command line:

openclaw pairing approve telegram <code>

7) Optional: Connect WhatsApp

The template provides a WhatsApp channel config (with allowlisting) and uses openclaw channels login to produce a QR code you scan in WhatsApp. (Template Instructions )

Data Persistence and Backups

On Zeabur, treat persistence as non-negotiable.

The template states all data is stored under /home/node, including config, credentials, and workspace/memory files.

Back up after initial setup using Zeabur’s built-in backup feature or manual backup from the Files tab.

Security Notes

OpenClaw is powerful because it can be wired into real tools and accounts — so deployment security matters.

Practical guidelines:

• Prefer loopback + private access patterns when possible. The official remote access docs describe keeping the Gateway bound to loopback (default) and using SSH tunnels or tailnets/VPNs for remote control. (Remote Access Guide )
• If you must expose access publicly, lock down who can reach it. OpenClaw’s security guidance recommends firewalling with tight allowlists when binding to LAN and warns against exposing the Gateway unauthenticated. (Security Guide )
• Run openclaw security audit as a quick health check — the docs frame this as a “how to not get pwned” workflow focused on access control and blast radius.

Zeabur-Specific Tip: Firewall Rules on Dedicated Servers

Zeabur offers firewall management for Dedicated Servers with controls like restricting source IP ranges. If you’re using a Dedicated Server deployment, that’s a strong lever for protecting an exposed Control UI. (Zeabur Firewall Changelog )

Alternative: Deploy as a Custom Docker Image

If you want full control (pin image tags, customize startup, add volumes), Zeabur supports deploying custom Docker images through “Add Service → Docker Images” and configuring ports, volumes, and start commands. (Zeabur Custom Docker Docs )

This is useful when:

• You want a specific OpenClaw image tag
• You’re composing multiple services
• You want to add adjacent infrastructure (reverse proxy, WAF, private networking, etc.)

Quick Reference Links