Practical OpenClaw: Deploying Your Private AI Assistant on Proxmox LXC

calendar 2026-01-29 · 4 min read · HomeLab AI Proxmox LXC OpenClaw Moltbot Clawdbot Self-Hosted  ·
Share on: linkedin copy

In the AI era, what can we do besides chatting with ChatGPT? The answer is: Let AI actually do things for you.

This is a practical guide on deploying OpenClaw (formerly Moltbot / Clawdbot) on Proxmox LXC. No complex Docker orchestration, no expensive cloud servers—just a lightweight LXC container to own a 24/7 private assistant that can monitor stocks and write blogs for you.

I. Why Bother?

There are many AI tools out there, but most are just "chatbots". You ask, they answer, and that's it. OpenClaw is different. It is an AI Agent.

Its Killer Features:

  1. Action-Oriented: It doesn't just chat; it uses tools (CLI, API, Browser).
  2. Privacy First: Self-hosted on your Proxmox, you control your data.
  3. Multi-Modal: Supports connections via Telegram, Slack, WhatsApp, etc.

My Real-World Use Cases

To prove it's not just a toy, here are two scenarios I use daily:

Scenario 1: Financial Analyst 💰

I set up a cron job for OpenClaw. Every day before the market closes at noon:

  • Auto-Check: Real-time prices of VOO (US) and VFV (Canada).
  • Read: Access my local cost basis file.
  • Analyze: Evaluate today's trend against my DCA strategy and message me on Telegram: "Market dip today, recommend buying $500" or "Too high, wait."
  • Result: I no longer need to check the market secretly during work.

Scenario 2: Auto-Blogging Assistant ✍️

The article you are reading right now was largely written by OpenClaw.

  • I gave it access to my GitHub Repo.
  • I provided a topic and a reference link.
  • It analyzed my writing style, wrote the Markdown, and submitted a Pull Request.
  • I just reviewed and merged it.

II. Environment Preparation

We will use Proxmox LXC, which is lighter than a VM and behaves more like a full OS than Docker, making it easier to debug.

Hardware Requirements

  • Platform: Proxmox VE (PVE)
  • Resources:
    • CPU: 2 Cores (Recommended)
    • RAM: 2GB - 4GB (More if running local Ollama)
    • Disk: 10GB+

III. Step-by-Step Deployment Guide

We will manually create a standard Ubuntu LXC (Proxmox Community Scripts work similarly).

Step 1: Create LXC Container

In Proxmox UI:

  1. Click Create CT.
  2. Template: Select ubuntu-22.04-standard.
  3. Resources: 2 Core, 2GB RAM, 8GB Disk.
  4. Network: Static IP recommended for easier SSH access.
  5. Unprivileged: Uncheck (Privileged containers are easier for certain mount scenarios, though unprivileged is safer if you know how to configure it).

Start the container and open Console.

Step 2: Initialize Environment

OpenClaw relies on Bun (a fast JS runtime).

1# Update System
2apt update && apt upgrade -y
3apt install -y curl unzip git
4
5# Install Bun
6curl -fsSL https://bun.sh/install | bash
7
8# Configure Environment Variables
9source /root/.bashrc

Step 3: Install OpenClaw

Use the official one-click installer:

1curl -fsSL https://openclaw.ai/install.sh | bash

Once installed, run the onboarding wizard:

1openclaw onboard

Step 4: Onboarding Configuration

The wizard will guide you through:

  1. Model Provider:

    • OpenAI / Anthropic: Requires API Key (Recommended for stability and smarts).
    • Ollama: Use local Ollama address if you have a GPU (Free, but power-hungry).
    • I use Google Gemini Pro via API, which offers great value.

  2. Channel:

    • Telegram is recommended.
    • Get a Bot Token from @BotFather on Telegram.
    • Enter it in the wizard.

Step 5: Verification

After installation, the OpenClaw Gateway service starts automatically. Check status with the TUI:

1openclaw tui

Or send Hello to your bot on Telegram. If it replies, congratulations! Your private Jarvis is online. 🎉

IV. Advanced Configuration: Persistence

1. Process Guard

Ensure OpenClaw restarts automatically after LXC reboot. Check systemd status:

1systemctl --user status openclaw-gateway

Note: OpenClaw registers as a user-level service by default.

2. Mount Storage (Optional)

If you want its "Memory" (Memory.md) and config to survive container destruction, use Proxmox Bind Mounts to map a host directory into the LXC.

V. Tips & Pitfalls ⚠️

  1. Network: Ensure your LXC can access GitHub/npm. Use a proxy if necessary.
  2. API Costs: OpenClaw is free, but OpenAI/Claude APIs are pay-per-use. Set usage limits.
  3. Security: Do not expose the Dashboard port (default 18789) to the public internet! Use Tailscale or VPN.

Conclusion

Deploying OpenClaw is just the beginning. The real fun is training it. You can teach it to write code, manage schedules, or reply to emails.

It's not perfect, but it's yours. In an era of closed AI ecosystems, owning an AI Agent on your own server is a kind of cyberpunk romance.

Happy Hacking! 🚀

comments powered by Disqus

Translations: