Here’s the thing about building software in 2026 – the rules have changed completely.
Last week, I shipped SiteKit, a full-featured server management platform built with Laravel 12 and Filament 3. It provisions servers, deploys applications, manages databases, handles SSL certificates, runs cron jobs, and monitors server health. The kind of product that would have taken months to build just a couple of years ago.
I built it in 7 days.
The Problem I Was Solving
If you’ve ever tried to deploy a Laravel application on a VPS, you know the pain. You either pay $12/month per server for Laravel Forge (which adds up quickly if you’re managing multiple servers), or you spend hours manually configuring Nginx, PHP-FPM, MySQL, Redis, and everything else.
I wanted something in between – a self-hosted solution that gives you the convenience of Forge without the recurring costs.
Why AI-Assisted Coding Changes Everything
Let me be clear about what “AI-assisted” means here. I didn’t just paste “build me a server management platform” into ChatGPT and call it a day. That’s not how this works.
What I did was use Claude as a pair programming partner. I would describe what I needed, discuss architecture decisions, and then implement the code with AI helping me write boilerplate, catch edge cases, and suggest improvements.
The real productivity gain isn’t in code generation. It’s in three areas:
1. Faster Decision Making
When you’re building something complex, you constantly face decisions – which database structure to use, how to handle job queues, what the API contract should look like. Normally, you’d spend time researching, reading documentation, and weighing options.
With AI assistance, I could describe my requirements and get well-reasoned recommendations instantly. Not always perfect recommendations, but a solid starting point that saved hours of research.
2. Boilerplate Elimination
Server management involves a lot of repetitive patterns. You need Filament resources for servers, web apps, databases, cron jobs, firewall rules, SSL certificates, and more. Each one follows similar patterns but with different fields and relationships.
I could describe the entity once and get a properly structured resource with forms, tables, and actions. Then I’d review it, make adjustments, and move on.
3. Error Handling and Edge Cases
This is where AI really shines. When I’m deep in implementation mode, I tend to focus on the happy path. The AI would often catch things I missed – “what happens if the SSH connection times out?”, “should we validate the domain format before creating an SSL certificate?”, “what if the user tries to delete a server that’s still provisioning?”
The Architecture
SiteKit has two main components:
The Laravel Application – Built with Filament 3 for the admin panel. It handles user management, team permissions, job queuing, and all the business logic. The UI is clean and functional – nothing fancy, just what you need to manage servers effectively.
The Go Agent – A small binary that runs on each managed server. It receives commands from the Laravel app via a secure API, executes them, and reports back. I chose Go for this because it compiles to a single binary with no dependencies – you just copy it to the server and run it.
The communication between them uses encrypted payloads and webhook callbacks. When you deploy an app, the Laravel side creates a job, sends it to the agent, and the agent reports progress through callbacks. This means the UI stays responsive even during long-running operations.
What SiteKit Actually Does
Here’s the feature list:
- Server Provisioning – Connect via SSH, install and configure all required services (Nginx, PHP-FPM, MySQL, Redis, Supervisor, etc.)
- Web App Management – Create sites with proper Nginx configurations, handle PHP and Node.js applications
- Database Management – Create MySQL/MariaDB databases and users with proper permissions
- SSL Certificates – Automated Let’s Encrypt certificate provisioning and renewal
- Cron Jobs – Manage scheduled tasks with a clean UI
- Firewall Rules – UFW management with an intuitive interface
- SSH Key Management – Add and remove SSH keys for server access
- Service Monitoring – Track service status and restart services when needed
- Deployments – Git-based deployments with hooks, shared files/directories, and rollback support
- Health Monitoring – Server resource monitoring with alerts
The 7-Day Timeline
Days 1-2: Core architecture – database schema, basic Filament resources, the Go agent structure, and the communication protocol between them.
Days 3-4: Server provisioning – this was the most complex part. Getting SSH connections working reliably, handling different Ubuntu versions, installing packages in the right order, and dealing with all the edge cases.
Days 5-6: App management and deployments – creating web apps, generating Nginx configs, setting up Git-based deployments with atomic releases.
Day 7: Polish and testing – SSL certificates, cron jobs, firewall rules, and lots of manual testing on real servers.
Lessons Learned
Start with the data model. I spent almost a full day just on the database schema before writing any application code. This paid off massively later – when your relationships are right, everything else flows naturally.
Build the agent first. Having a working agent that could execute commands made it easy to test the Laravel side. I could verify each feature worked on real servers as I built it.
Don’t over-engineer. The temptation with a project like this is to make everything configurable and handle every edge case. I deliberately kept things simple – support Ubuntu, support the most common setups, and make it work well for those cases.
AI is a tool, not a replacement. The AI made me faster, but I still had to understand what I was building. When something went wrong (and things always go wrong), I had to debug it myself. The AI couldn’t SSH into my test server and figure out why Nginx wasn’t starting.
What’s Next
SiteKit is now open source under a source-available license. You can self-host it, modify it, and use it for your own servers. The only restriction is that you can’t offer it as a hosted service to compete directly with the project.
I’m continuing to add features – Node.js support is already in, and I’m working on better monitoring and alerting. But the core product is solid and I’m already using it to manage my own servers.
The Broader Point
What took me 7 days would have taken 3-4 months a few years ago. That’s not because I’m some exceptional developer – I’m decent, but not special. It’s because the tools have gotten that much better.
This doesn’t mean AI is going to replace developers. It means the bar for what a single developer can build has risen dramatically. The same way cloud infrastructure let startups do things that previously required a data center, AI assistance lets individual developers build products that previously required teams.
If you’re a developer and you’re not experimenting with AI-assisted coding, you’re leaving a lot of productivity on the table. It’s not about replacing your skills – it’s about amplifying them.
SiteKit is available at https://sitekit.dev – You can check out the source code, deploy it on your own server, and start managing your infrastructure without monthly fees.