When Apple discontinued macOS Server.app, Mac users lost the simple, point-and-click way to run websites and services. FrontierStack brings that back and goes much further: it installs, controls, secures and monitors Apache, PHP, MySQL, PostgreSQL, Nginx and more than a thousand other services — on the Mac in front of you and on every server you link to it — and it puts a tool-using AI administrator at the helm.

This manual is the complete reference for the desktop app. A companion Mobile Manual covers the iPhone and iPad app. Read this first chapter end to end; after that, dip into whichever part you need.

1.1What FrontierStack is

FrontierStack is a native macOS application — not a web panel, not a container you have to babysit. It manages real services on real machines using the tools they already ship with: Homebrew formulae, launchd and systemd units, Apache and Nginx config, certificates, firewalls and SSH. It is three things at once:

• A control panel. Install, start, stop and configure services with validation, live status and one-click fixes — the macOS Server replacement, modernised.
• A monitoring and alerting system. A continuous health sweep watches services, sites, certificates, disks and your whole fleet, and turns the first sign of trouble into a message on the channels you choose.
• An AI administrator. A guard-railed, tool-using assistant with a live, accurate picture of your setup. It can diagnose what is broken, propose a fix you approve, run it locally or across your fleet, and reshape the app to exactly what you run.

1.2An AI-first design

Most apps bolt a chat box onto the side. FrontierStack is built the other way around: the AI is a first-class operator of the same controls you use. Because it acts only through a fixed set of audited tools — read a log, run a diagnostic, restart a service, issue a certificate, write a file — everything it does is something you could do by hand, and everything that changes the system is shown to you for approval first.

Three principles make this safe enough to trust with production servers:

• Read-only by default. Out of the box the assistant can look but not touch. Changes and scripts are unlocked by explicit toggles, and every mutating action shows an approval card before it runs.
• Your secrets never reach the model. Passwords and API tokens live in a local .env vault. Scripts reference them by name; the app injects the real value at run time, locally or over SSH. The cloud model sees the name, never the value. Logs and files are secret-redacted before they are sent.
• You bring your own AI. Point it at a cloud model or a local one (Ollama, LM Studio). If you drive it from your own subscription tool over MCP — Claude Code, Cursor — there is no metered per-token bill at all.

1.3Who it is for

FrontierStack suits anyone running — or wanting to run — servers on a Mac:

• Developers who want a clean local stack instead of MAMP or XAMPP.
• Small businesses self-hosting POS, CRM, e-commerce, mail and helpdesk on a Mac mini.
• Admins keeping a vintage Xserve or Mac mini on macOS Server useful by managing it remotely.
• Anyone moving from metered SaaS to self-hosted, or running a fleet of Macs, Linux boxes, BSD servers and Raspberry Pis from one window.

1.4Key concepts and vocabulary

A handful of terms recur throughout this manual. Meet them once here.

1.5How this manual is organised

The manual follows the shape of the work:

• Part I — Getting Started: install, first run, and a tour of the interface.
• Part II — Running Services: the web stack, databases, the full catalog, and sites & domains.
• Part III — Fleet, Network & Security: linking servers, networking and the perimeter, hardening, and monitoring & alerts.
• Part IV — Automation & AI: scripts and schedules, the AI Administrator in full, MCP, and the AI stack.
• Part V — Business & Reference: vertical solutions, settings, shortcuts, licensing and troubleshooting.

If you are brand new, continue to Chapter 2, Installation & First Run. If you just want to understand the AI, jump ahead to Chapter 13 — but the safety model there assumes the concepts above.