What is ServerGuardAI?

ServerGuardAI keeps an eye on your Linux servers 24/7 and sends you instant alerts when something goes wrong. Check server health, restart services, block suspicious IPs, and get AI-powered diagnostics, all from your device.

Agent

A tiny program on your server that sends health data every 2 minutes.

Cloud API

Receives the data, stores it, checks thresholds, and sends you push notifications.

The App

Your dashboard to see everything and take action when needed.

1

Add Your Server in Seconds

Getting started takes just a few taps. Enter your server’s hostname or IP address and your SSH login details (password or SSH key).

The app connects to your server via SSH, automatically installs the monitoring agent, and starts it as a background service. No terminal commands needed.

  • Automatic agent installation via SSH
  • No manual terminal commands required
  • Agent starts as a systemd service, auto-starts on boot
  • Works on Ubuntu, Debian, CentOS, AlmaLinux, Rocky Linux, and more
  • Keeps itself up to date with signed releases, so you never SSH in to update it again
Server list in ServerGuardAI app
2

Your Dashboard, Updated Live

Within seconds of adding your server, live metrics start flowing to your dashboard. Everything you need to know at a glance, updated every 2 minutes.

  • CPU & Memory: Current usage with visual gauges
  • Disk usage: How full each partition is
  • Load average: How busy the server is overall
  • Top processes: Which programs consume the most resources
  • Services: Nginx, Apache, MySQL, PHP-FPM, Redis, and more
  • SSL Certificates: Auto-discovered domain certificates and expiry dates

Start, stop, or restart any service with a single tap, commands run directly on your server via SSH.

Server detail page with metrics
3

Instant Alerts When It Matters

Get notified the moment your server needs attention. Available on all plans, Free gets 100 push alerts, Pro and Agency get unlimited.

  • CPU above 90%: Server is under heavy load
  • Memory above 90%: Running low on RAM
  • Disk above 95%: Drive is almost full
  • High load average: Too many tasks waiting to run
  • SSL expiring soon: Certificate needs renewal
  • Service down: A critical service has stopped
  • Fatal errors in logs: Critical/fatal entries detected in server logs
  • Login after repeated failures: Someone got in right after a burst of failed attempts
  • PHP pool down: The site still loads from cache, but PHP behind it is dead
  • Nightly check didn’t run: Silence is reported, not assumed to be good news

Alerts are delivered as push notifications to all your devices. Repeat alerts for the same problem are throttled per alert type, from 5 minutes for crash loops to 24 hours for slow-moving conditions like disk space, so you don’t get spammed. Snooze or mute any alert type per server, set quiet hours, and get a weekly summary you can open and read instead of a notification that goes nowhere.

Alerts view in ServerGuardAI
4

AI-Powered Diagnostics

Available on all plans, each analysis uses 1 AI credit. Two ways to ask. Tap Diagnose on an alert and you get one answer: what caused it, the evidence, and the one command that fixes it, with a plain verdict like “flat for a week, not urgent tonight”. Or tap Analyse on a server and the AI examines your metrics, logs, service status, and network connections, then gives you a ranked list you can work through:

  • Performance issues: CPU bottlenecks, memory pressure, slow queries
  • Security threats: Brute-force attacks, suspicious IPs, exposed ports
  • Database problems: Slow queries, connection issues, misconfiguration
  • Network anomalies: Unusual traffic patterns, DDoS indicators
  • Storage concerns: Disk space, large files, log rotation

Each finding includes a confidence score, the root cause, supporting evidence, and a recommended fix with one-tap execution. The AI also provides verify and rollback commands. You always review the command first, nothing runs without your approval.

Disk full? You get a button that frees the space and tells you how much, not ten lines of advice. It only ever touches logs, journals, caches and crash dumps, never your sites or data, and it says so.

Findings you mark as fixed, expected, or not a real issue stay that way. Free gets 3 credits, Pro gets 10, Agency gets 75, all one-time, not monthly. Additional packs of 50 credits can be purchased anytime on any plan.

AI diagnostics in ServerGuardAI
5

Manage & Protect Your Servers

Take action on your servers without ever opening a terminal. All commands run directly from your device to your server over SSH.

  • Restart services: Nginx, Apache, MySQL, PHP-FPM, and more
  • Block & unblock IPs: Works with iptables, UFW, firewalld, and nftables
  • AI-recommended blocks: Block suspicious IPs flagged during analysis
  • Kill processes: Stop runaway processes consuming resources
  • Update or remove the agent: Full control over the agent lifecycle

Built-in safety: Before any command runs, a safety filter blocks dangerous operations like recursive deletes, system shutdowns, and fork bombs. Every action is logged so you have a complete audit trail.

AI report with recommendations
6

Checked Every Night, From Outside Your Website

Available on every plan, including Free. Switch it on per server and, every night, the agent checks each website on that server from outside the site itself, where a compromised plugin can’t switch it off or lie to it.

  • Malware scan: Web shells, backdoors, injected scripts, and files that have no business being in your site folders
  • Exposed files, proven not guessed: Backups, database dumps, .env files and .git folders are checked by actually fetching them from the internet, and only what could be reached is reported
  • Admin tools left open: Database consoles and file managers reachable by anyone on the internet
  • WordPress plugin and theme inventory: Everything installed, its version, what is out of date, and whether core and plugin files match the official releases
  • Hidden WordPress administrator accounts: Accounts that exist in the database but are hidden from the admin screen
  • Updates that don’t bury you: An update that touches 400 files shows up as one line, not 400 alerts

The malware scan, the exposed-file checks and the open admin-tool check run on every PHP site. Plugin inventory, release comparison and hidden-administrator detection are WordPress-specific today; Laravel, Symfony and CodeIgniter equivalents are on the roadmap.

Every scan leaves a receipt. You see what was checked, how many files, and when, not just that nothing was found. If a check didn’t run, you hear about that too. Fix something and the finding closes itself once the next check confirms it. Dismiss something and it stops asking.

The check runs at the lowest priority with a memory cap, at a random time in the early hours. On a 1-core test server with a full WordPress site it finished in about 90 seconds.

Nightly check findings in ServerGuardAI

How Data Flows

Two separate channels: monitoring goes through the cloud, actions go direct.

Monitoring

Your Server
Agent collects metrics
↓ Encrypted HTTPS
ServerGuard Cloud
Stores & checks thresholds
↓ Push notification
Your Device
Dashboard updates live

Actions

Your Device
You tap an action
↓ Direct SSH
Your Server
Command runs immediately

The cloud API is never involved in actions, it only handles monitoring data and notifications.

Built for Security

Security is designed into how the pieces talk to each other.

One-Way Agent

The agent only sends data out. It never receives or runs commands from the cloud. No open ports needed on your server.

Encrypted & Signed

All data travels over HTTPS with TLS. Everything the agent sends is additionally signed with HMAC-SHA256 to prevent tampering.

Keychain Credentials

SSH passwords and keys are stored in Apple’s Keychain, backed by hardware security. They never leave your device.

Command Safety Filter

Destructive commands (recursive deletes, shutdowns, fork bombs) are automatically blocked. Each server has isolated API credentials generated at setup.

Signed Agent Updates

Every agent release is cryptographically signed. The agent verifies the signature before installing, refuses anything unsigned or older than what it runs, rolls out gradually, and rolls back on its own if a release misbehaves.

Checks From Outside the Site

The nightly check reads your site’s files from the operating system, not from inside the website’s own code. Something running inside a compromised site can’t switch it off or feed it false answers, and the receipt shows exactly what was looked at.

Works Across All Your Devices

Use ServerGuardAI on your iPhone, iPad, and Mac. Server configurations can be synced across devices via iCloud (push/pull from Settings). All devices see the same live metrics and alerts from the cloud API. SSH credentials are stored securely in Keychain on each device and must be entered once per device.

Plans & Pricing

The same protection on every plan. Tiers differ by capacity.

Free

$0/month
  • 1 server
  • 100 push alerts
  • 3 AI credits (one-time)
  • Malware scanning
  • AI diagnostics
  • Buy AI credit packs
Get Started
Full Power

Agency

$19.99/month
  • Up to 20 servers
  • Unlimited push alerts
  • 75 AI credits (one-time)
  • Malware scanning
  • AI diagnostics
  • Buy AI credit packs
Start Agency

Additional AI credit packs (50 credits) can be purchased anytime on any plan.

Questions Developers Ask

The things you would want to know before letting anything run as root on a server you are responsible for.

What exactly runs on my server, and as which user?

A bash agent as a systemd service, running as root because it needs to read every service state and log. It is capped at 128 MB and wakes every 2 minutes. If you turn on nightly checks, a second unit runs once a night at the lowest CPU priority with a 512 MB memory cap and a 4-hour hard stop. Both are plain bash you can read at /usr/local/bin before and after install. Nothing listens on a port.

Can your cloud send commands to my server?

No. The agent only pushes data out. Restarting a service or blocking an IP goes from your device straight to your server over SSH. Our API never sees the command and has no path to your machine, even if it were compromised.

What data leaves my server?

Every 2 minutes: CPU, memory, disk, load, service states, SSL expiry dates, the top IPs and error lines from your web logs, and the last 300 lines of each error and access log (Pro and Agency). Emails, passwords, tokens and auth headers are scrubbed from log lines on arrival. Nightly checks send file paths, sizes and hashes, which rule matched, and plugin names and versions. Your files are never uploaded.

How long do you keep it?

Raw metrics 6 hours, 10-minute averages 3 days, hourly averages 90 days. Alerts 30 days. Log snapshots are overwritten every 2 minutes, so only the newest copy ever exists. Deleting a server in the app deletes everything stored for it.

Does the installer put anything else on my box?

curl, jq, openssl and awk if they are missing, through your package manager, plus a small signature-verification binary for updates. It never installs wp-cli or anything inside your websites, and it never changes your firewall.

Does the agent update itself, and can I stop that?

Yes, from signed releases only. It verifies the signature, refuses anything unsigned or older than what it runs, rolls out gradually and rolls back on failure. Switch it off per server in the Security screen and update by hand whenever you prefer.

Will the nightly check slow down a small VPS?

Measured on a 1-core, 1 GB server running a full WordPress site: about 90 seconds, under 10 MB of memory, and page response times went from under a millisecond to a few milliseconds while it ran, then back. It runs at the lowest CPU priority, is memory-capped, and is off by default until you turn it on for that server.

What if it flags something that is mine?

Mark it as expected and it never asks again, and you can undo that any time. The detection rules were measured against 815,000 PHP files on 103 production sites before shipping, and five plausible-looking rules were deleted for false positives. Findings that cannot be confirmed are shown as warnings and never send a push.

Does it work in Docker, behind Cloudflare, or behind NAT?

For monitoring, yes: the agent only makes outbound HTTPS connections, so NAT, Cloudflare and firewalls do not get in the way. Inside a container it reports host-level metrics. For actions, your device needs to reach the server over SSH directly. Jump hosts are not supported yet.

Which distros and architectures?

Ubuntu, Debian, AlmaLinux, Rocky, CentOS and Amazon Linux 2023, on x86_64 and arm64. Older systems such as Amazon Linux 2 run the monitoring agent but cannot run the nightly check. Linux only, no Windows.

Can I change the alert thresholds?

Thresholds are fixed and tuned against real servers to avoid noise: CPU, memory and load need three consecutive breaches over six minutes before anything fires. You can mute or snooze any alert type per server, set quiet hours, and choose which severity levels are allowed to push.

What happens if your API goes down?

The agent queues metrics and alerts locally and sends them when it can reach us again, so history fills back in. Everything you do from the app over SSH is unaffected. We never report our own outage as if it were your server going down.

Can my team use it? Android? Web?

Anyone with SSH access to a server can add it and receive the same alerts alongside you. iPhone, iPad and Mac today; there is no Android or web version. Read-only sharing by invite code and Slack delivery are the next release.

Can I read the code before I trust it?

The agent, the nightly scanner and the installer are plain bash. They sit on your server at /usr/local/bin after install, and the identical files are served from api.serverguardai.com/downloads so you can read them first. The app and the API are not open source.

Get Started in Under 2 Minutes

Download the app, add your server, and see your first dashboard, the agent installs automatically.

Download ServerGuardAI

Common Questions

Everything you need to know before getting started.

Any Linux server you can SSH into, Ubuntu, Debian, CentOS, AlmaLinux, Rocky Linux, and more. The agent is a lightweight bash script with no special dependencies.

No. The agent uses minimal resources, it runs for a fraction of a second every 2 minutes to collect metrics, then sleeps. It’s designed to be invisible.

Automatically. When you add a server in the app, it connects via SSH, installs the monitoring agent, and starts it as a background service. No terminal commands needed.

Yes. Anyone with SSH access to a server can claim it using the server’s API key. The original owner keeps full ownership, and the new user gets shared access to view metrics and alerts.

Yes. Monitoring data (CPU, memory, etc.) is stored securely on our cloud API. SSH credentials never leave your device, they’re stored in Apple’s Keychain, backed by hardware security, and are never sent to our servers.

The agent continues running and sending data. If you reinstall the app, your servers will sync back from iCloud. You can also remove the agent from any server at any time via SSH.