Deploying OpenClaw(Moltbot/Clawdbot) on Proxmox

📢 Update: February 2, 2026

The project has rebranded again to OpenClaw (from Moltbot/Clawdbot) to emphasize its open-source nature and avoid further trademark issues. This is the final name—commands are now openclaw instead of clawdbot. Installer and docs links have been updated below. The core functionality remains the same, with added security hardening (34+ commits), new model support (e.g., KIMI K2.5, Xiaomi MiMo-V2-Flash), and channels like Twitch and Google Chat. Old names like Moltbot/Clawdbot are kept in searches for compatibility.

I’ve been watching the AI assistant space for a while now, and most solutions are either cloud-only, expensive, or just don’t do enough. When I discovered OpenClaw (formerly Moltbot/Clawdbot), I was intrigued—here’s an open-source AI assistant that actually performs actions, not just answers questions. It clears your inbox, manages your calendar, sends emails, and integrates with WhatsApp, Telegram, and Slack. Best of all? You can self-host it.

After deploying it on my Proxmox homelab, I wanted to share what makes OpenClaw worth the setup effort and how to get it running efficiently.

Here’s what I’ll cover:

• What OpenClaw is and why it’s different
• Key features that make it worth self-hosting
• Understanding Moltbook and the agent ecosystem
• Deploying on Proxmox using the official installer
• Basic Proxmox setup considerations
• Real-world use cases and limitations

What Makes OpenClaw Different?

Most AI assistants are glorified chatbots. They answer questions, maybe help with writing, but they don’t actually do things. OpenClaw is different—it’s designed to be an AI agent that performs actions across your digital life.

Core Capabilities

• Email Management: Clears inboxes, sends emails, organizes messages
• Calendar Automation: Manages your schedule, books meetings, sends reminders
• Task Management: Creates and tracks tasks across platforms
• Multi-Platform Integration: Works through WhatsApp, Telegram, Slack, or any chat app (new: Twitch, Google Chat)
• Workflow Engine: Features “Lobster,” a workflow system for composable automations
• Persistent Memory: Remembers context and preferences across conversations

Understanding Moltbook and the Agent Ecosystem

Moltbook is a Reddit-style social network launched in January 2026 exclusively for AI agents. Think of it as “the front page of the agent internet”—only verified AI agents (primarily running OpenClaw) can post and interact, while humans can only observe.

Why Moltbook Matters

Moltbook serves as:

• Skills Marketplace: Agents share and discover new capabilities (called “skills”)
• Agent-to-Agent Communication: Your OpenClaw instance can interact with 770,000+ other agents
• Community Learning: Agents collaborate on problem-solving and share workflows

The Moltbook Security Incident (January 31, 2026)

Supply Chain Risks

Beyond the database exposure, Moltbook introduces supply chain risks:

• Malicious Skills: Researchers found skills like “What Would Elon Do?” that contained active data exfiltration code
• Prompt Injection: Agents can embed malicious instructions in posts that manipulate other agents’ behavior
• Delayed Execution Attacks: Thanks to persistent memory, malicious payloads can be fragmented across multiple posts and assembled later

Cybersecurity firms like Cisco and 1Password have documented these risks extensively, noting that OpenClaw agents with Moltbook access often run with elevated permissions on users’ local machines.

Why Deploy on Proxmox?

Proxmox is perfect for this because:

• Isolation: Run OpenClaw in a dedicated VM or LXC container
• Resource Management: Easy to allocate CPU, RAM, and storage
• Backup Integration: Leverage Proxmox Backup Server for automated backups
• Network Control: Configure firewall rules and network isolation
• Snapshots: Test configurations without breaking production

I’m running OpenClaw in an Ubuntu 22.04 LXC container with 4GB RAM and 2 CPU cores, which handles my personal use case perfectly. For heavier workloads or multiple users, you might want to allocate more resources (e.g., 8GB+ for Ollama models).

Prerequisites

Before we start, make sure you have:

• Proxmox installed and accessible
• Basic networking configured (static IP recommended)
• An AI model provider (OpenAI API, Anthropic Claude, or self-hosted Ollama; new support for KIMI K2.5 & Xiaomi MiMo-V2-Flash)
• Node.js/Bun runtime (OpenClaw runs on Bun or Node.js)

Setting Up the Container

Create an Ubuntu 22.04 LXC container in Proxmox with:

• Disk: 20GB
• CPU: 2 cores
• Memory: 4096 MB
• Network: Static IP or DHCP

Start the container and SSH into it.

Step 2: Install Bun Runtime

OpenClaw runs on Bun (a fast JavaScript runtime). Install it in your container:

# Update system
apt update && apt upgrade -y

# Install required dependencies
apt install -y curl unzip

# Install Bun
curl -fsSL https://bun.sh/install | bash

# Add Bun to PATH (add to ~/.bashrc or ~/.zshrc)
export BUN_INSTALL="$HOME/.bun"
export PATH="$BUN_INSTALL/bin:$PATH"

# Verify installation
bun --version

Installing OpenClaw

Option 1: Using the Official Installer

The official installer is the recommended method. Run the installer script:

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

This installer will:

• Install OpenClaw globally via npm (requires root/sudo)
• Set up the necessary configuration directories
• Make the openclaw command available system-wide

For non-root installations, use the CLI-only installer:

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

After installation, the onboarding wizard starts automatically. If it doesn’t, run:

openclaw onboard

Configuration

After installation, the onboarding wizard starts automatically. If it doesn’t, run:

openclaw onboard

Onboarding Process

The wizard guides you through:

1. Security Warning: Read and acknowledge the security notice
2. Onboarding Mode: Choose QuickStart (recommended)
3. Model Provider:
   • OpenAI: Uses OAuth (browser opens automatically)
   • Anthropic: API key
   • Ollama: Self-hosted URL (e.g., http://your-ollama-server:11434)
4. Channel Setup: Choose your messaging platform:
   • Telegram: Get bot token from @BotFather (/newbot)
   • WhatsApp: QR code pairing
   • Discord: Bot token
   • Slack: Socket Mode token
5. Skills: Configure optional skills (skip for now)
6. Hooks: Enable automation hooks (optional)
7. Agent Setup: Configure your first agent in the TUI

For Telegram (Easiest)

1. Open Telegram and message @BotFather
2. Run /newbot and follow instructions
3. Copy the bot token (format: 123456:ABC...)
4. Paste it when prompted in the wizard

Starting OpenClaw

After onboarding, the Gateway service starts automatically. Access:

• TUI: openclaw tui (recommended)
• Web UI: openclaw dashboard (opens browser)
• Check status: openclaw status

Send a message to your Telegram bot (or other configured channel) to start chatting.

Proxmox Security Hardening

Given the security risks outlined above, here are essential hardening steps for Proxmox deployments:

Network Isolation

# In Proxmox, create a dedicated VLAN for OpenClaw
# Firewall rules (example for iptables in container):

# Allow only outbound connections
iptables -A OUTPUT -m state --state ESTABLISHED,RELATED -j ACCEPT
iptables -A OUTPUT -p tcp --dport 443 -j ACCEPT  # HTTPS only
iptables -A OUTPUT -p tcp --dport 80 -j ACCEPT   # HTTP (if needed)
iptables -P OUTPUT DROP

# Block all inbound except from Proxmox host
iptables -A INPUT -s YOUR_PROXMOX_IP -j ACCEPT
iptables -A INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT
iptables -P INPUT DROP

Unprivileged LXC Container

Create the container as unprivileged to prevent root escalation:

# In Proxmox:
# When creating LXC, uncheck "Privileged container"
# This maps container root to unprivileged host UID

Dashboard Access

Never expose the dashboard to the public internet. Use one of these methods:

1. Tailscale VPN: Install Tailscale in the container for secure remote access
2. SSH Tunnel: ssh -L 1455:localhost:1455 user@proxmox-host
3. Reverse Proxy with Auth: Use Nginx with HTTP basic auth (only if necessary)

Credential Isolation

Use environment variables or secrets management instead of storing credentials in config files:

# Instead of hardcoding in openclaw.json:
export ANTHROPIC_API_KEY="your-key-here"
export OPENAI_API_KEY="your-key-here"

# Start gateway with env vars
openclaw gateway start

Monitoring and Auditing

Set up log monitoring to catch suspicious activity:

# Monitor gateway logs in real-time
tail -f ~/.openclaw/logs/gateway.log

# Set up alerts for specific patterns
grep -i "error\|unauthorized\|injection" ~/.openclaw/logs/gateway.log

Proxmox Notes

• Ensure container has outbound internet access for API calls
• Set up automated backups via Proxmox Backup (backup ~/.openclaw/ directory)
• Monitor resource usage through Proxmox web UI
• Consider using Proxmox firewall at the container level for additional security

What You Can Actually Do

Here are some real-world use cases I’ve been testing:

Email Management

• “Clear my inbox of newsletters”
• “Send an email to [person] about [topic]”
• “What emails need my attention today?”

Calendar Automation

• “Schedule a meeting with [person] next Tuesday at 2 PM”
• “What’s on my calendar this week?”
• “Move my 3 PM meeting to tomorrow”

Task Management

• “Create a task to review the quarterly report”
• “What tasks are due this week?”
• “Mark ‘buy groceries’ as complete”

Multi-Platform Access

The beauty of OpenClaw is that you can interact with it through WhatsApp, Telegram, or Slack—whatever you already use. No need to learn a new interface.

Troubleshooting

Gateway won’t start:

# Check logs
cat ~/.openclaw/logs/gateway.log

# Check status
openclaw status

# Restart gateway
openclaw gateway restart

Telegram bot not responding:

• Verify bot token is correct
• Ensure bot is started in Telegram (message /start to your bot)
• Check Gateway is running: openclaw status

OAuth issues (OpenAI):

• Ensure browser can access localhost:1455 for callback
• If callback fails, paste the redirect URL manually when prompted

Container networking:

• Verify outbound internet: ping 8.8.8.8
• Check Proxmox firewall allows outbound connections
• For Ollama: curl http://your-ollama-server:11434/api/tags

Prompt injection detection: If you suspect your agent has been compromised:

# Check for suspicious commands in logs
grep -E "rm -rf|curl.*sh|wget.*sh|/dev/null" ~/.openclaw/logs/gateway.log

# Review recent skill installations
openclaw skills list

# Reset API keys if compromised
# Then update in openclaw.json or re-run onboarding

Limitations and Considerations

OpenClaw is powerful, but it’s not perfect:

• API Costs: If using OpenAI or Anthropic APIs, costs can add up with heavy usage
• Learning Curve: Setting up integrations requires some technical knowledge
• Privacy: Even self-hosted, you’re still calling external APIs unless using Ollama
• Reliability: Self-hosted means you’re responsible for uptime and troubleshooting
• Platform Support: Some integrations may require additional setup or have limitations
• Resource Usage: Ollama models require significant RAM (8GB+ for larger models)
• Security Responsibility: You’re responsible for keeping credentials secure and the instance hardened

Next Steps

Once you have OpenClaw running:

1. Pair with Messaging Platforms: Connect WhatsApp, Telegram, or Slack
2. Configure Skills: Explore the skill directory for extensions (vet them carefully for security)
3. Set Up Workflows: Use Lobster to create custom automation pipelines
4. Monitor Usage: Track API costs and resource usage
5. Join the Community: Check out the OpenClaw GitHub for updates and support
6. Avoid Moltbook: Until security is verified, do not connect to Moltbook

Final Thoughts

OpenClaw isn’t just another AI chatbot—it’s a practical automation tool that actually does things. Deploying it on Proxmox gives you full control over your AI assistant while keeping costs predictable and data private.

The setup isn’t trivial, and the recent security incidents around Moltbook and exposed instances highlight that this technology requires careful deployment. However, if you follow proper security practices—network isolation, credential management, and monitoring—you can safely run OpenClaw in your homelab.

For homelab enthusiasts who understand the risks and are willing to invest time in proper security hardening, OpenClaw represents an exciting glimpse into the future of personal AI agents.

If you’re interested in seeing OpenClaw in action or have questions about the deployment, feel free to reach out. The community is active and helpful, and there’s plenty of documentation to get you started.

Resources

• OpenClaw GitHub Repository - Official source code and issues
• OpenClaw Documentation - Complete documentation site
• Getting Started Guide - Step-by-step setup
• Installation Options - All available installation methods
• Setup Guide - Post-installation configuration
• OpenClaw Website - Official website
• Proxmox Documentation - Proxmox VE documentation
• Bun Runtime Documentation - Bun runtime documentation
• Moltbook Security Analysis (404 Media) - Critical vulnerability report