Project Showcase

AI-Openclaw

AI-powered server management through messaging. Ask in natural language, get real answers from your infrastructure.

View on GitHub
"Why is the site slow?" → AI Agent → Checks CPU, RAM, nginx → "RAM at 94%, restarted php-fpm"
// How it works
You "Why is the website slow?" Discord / Telegram Messaging channel OpenClaw agent VPS (Ubuntu) — receives message, decides what to do api call Cloud LLM Via Ollama — analyzes request, checks CPU, RAM, nginx, connections tool call ssh Client server 1 checks nginx, RAM, CPU Client server 2 idle Client server 3 idle output AI analyzes output → replies in Discord "Nginx worker connections maxed out. RAM at 94%. Restarted php-fpm." 1 2 3 4 5 6 1. You ask a question 2. Message received 3. Agent processes 4. LLM decides commands 5. SSH executes on server 6. AI analyzes & replies
// Architecture
💬

Messaging Channels

Talk to your servers through Discord or Telegram. Just type a question in natural language — no SSH sessions, no memorizing commands.

Discord Telegram
messages
🧠

Central Agent Server

A dedicated VPS running the AI agent as a systemd service. Receives messages, orchestrates tool calls, manages SSH connections to all managed servers.

Ubuntu OpenClaw systemd
api
⚡

Cloud LLM via Ollama

AI model runs remotely through Ollama's cloud routing — zero local RAM usage. Supports tool calling so the agent decides which commands to run.

Free Tier Tool Calling 64k+ Context
ssh

Client Servers

Each server has a dedicated monitoring user with SSH key authentication and restricted sudo. The agent connects over SSH, runs commands, and analyzes results.

Web Server

cPanel / HestiaCP

App Server

Ubuntu

Dev Server

Ubuntu

+ more

Add anytime
// Automated monitoring

Health Checks & Alerts

Automated heartbeat checks run every 30 minutes across all servers. Alerts are posted to a dedicated Discord channel when thresholds are breached.

Disk Usage
Alert when > 90% full
Memory Usage
Alert when > 90% used
CPU Load
Alert when > 2x CPU count
Web Server Status
Alert if Nginx/Apache down
// Security model

Defence in Depth

Every layer is hardened. The monitoring user can manage services and read logs — but cannot destroy anything.

SSH Key Only
No passwords, custom ports
Restricted Sudo
rm, reboot, dd, mkfs, passwd blocked
No Root Access
Root login disabled everywhere
Firewall + Fail2ban
Brute force protection
AllowUsers Policy
Only authorized users can SSH in
Automated Provisioning
One-command server hardening
// Tech stack

OpenClaw

AI Agent Framework

Ollama

LLM Gateway

Discord

Chat Interface

SSH

Secure Connections

Bash

Provisioning Scripts

Node.js

Runtime