Lifestyle Portal Ajme Cuts Origin Traffic to 20GB of 147GB With Cloudways MCP Server

About Ajme

Ajme (ajme.hr) is a Croatian lifestyle portal based in Split, covering travel, events, culture, health, and everyday life across Dalmatia and beyond. The site runs on WordPress under Ajme portal j.d.o.o. and has grown to around 1,100 published articles, recently adding a curated Split events calendar.

Behind the editorial team sits one engineer, Ante Projić, who joined the portal shortly after its move to Cloudways. Rather than writing content, Ante manages everything that keeps the publication running: hosting, performance, security hardening, GDPR compliance, email deliverability, and custom WordPress development.

Ajme now manages its infrastructure through Cloudways MCP Server, using AI assistants to monitor performance, investigate traffic, and handle routine server operations without leaving the conversation. Here’s how one engineer keeps a growing publication running smoothly, without ever slowing the team down.

The Challenge: A Fragile Inheritance

Ante Projić arrived at Ajme after the migration to Cloudways had already happened. That meant the site’s history was legible mostly in what the move had left behind.

Old habits from a shared-hosting era lingered in the codebase. As Ante Projić described it, “I actually inherited the site just after the move to Cloudways, so I saw the ‘before’ mostly through what the migration left behind.”

“A shared-hosting mindset, plugin sprawl, no staging environment, and no clear separation between caching layers.” — Ante Projić

None of that sits well with a small team publishing across travel, events, health, and DIY lifestyle content, with no dedicated ops staff to catch problems before they surface.

– Mobile views of the Ajme homepage, showing the site’s lifestyle, travel, and events sections displayed across a clean, image-forward layout. (Source: ajme.hr)

Untested changes could go live with no safety net. Caching layers that should have worked together instead pulled against each other. A broken deploy had no clean way to be isolated before it reached readers.

All of it landed on one engineer, working largely alone.

The Solution: MCP Runs The Backend

The one thing that didn’t need fixing was the server itself. From day one, the stack, Varnish, Redis, Nginx/Apache, PHP versions, SSL, was already sound, which meant Ante Projić’s attention could go to the application layer instead of the operating system underneath it.

On that foundation, a deliberately simple workflow took shape: WordPress with the block editor for writing, and a heavily customized front end underneath it.

Where Ajme’s range of topics might have tempted a stack of plugins, small custom modules went in instead: redesigned category archives with per-topic SEO, curated homepage sections, an events calendar with recurring dates and calendar exports. Editors self-serve through shortcodes and checkboxes. Nobody touches code.

A caching stack combining Varnish, Redis, and Cloudflare edge means a viral post or a summer traffic spike no longer dictates the publishing schedule. But when the team wanted to confirm that setup was actually holding, it was Cloudways MCP that supplied the proof.

“Through the MCP I pulled server settings, Cloudflare details, and traffic analytics.” — Ante Projić

Running underneath the daily workflow, MCP handles the quiet, repetitive work: DNS management and analytics queries save the most time of anything in the toolkit. When the team suspected the site was heading toward a bandwidth problem, that single MCP query settled it, confirming the edge cache was already carrying the load and image optimization was doing its job. A worry that would once have meant days of panel screenshots and spreadsheets was closed out in a single afternoon.

– A Cloudways MCP session in Cursor, showing a plain-language query for server CPU and memory usage returned as structured data. (Source: cloudways.com)

Cache purges, the most routine task of all, now take one line after every deployment. Ante Projić put it simply: “Cache purges are the everyday one: after every deploy, one line, done.”

The Result: One Engineer, Real Scale

What emerged is a lean operation running at real scale on one part-time engineer’s time:

• Warm TTFB of 80 to 120 ms on a modest single Cloudways server
• Homepage HTML shipping at roughly 37 KB brotli-compressed, with about 91% of homepage images in WebP format
• Roughly 85% of monthly traffic served directly from the Cloudflare edge cache; the origin handling only about 20 GB of the site’s roughly 147 GB total egress
• Zero unplanned downtime since the migration
• Around 1,100 published articles running on a stack one part-time engineer maintains

“AI compresses the distance between ‘something’s off’ and ‘here’s the verified fix,’ and the publisher keeps the hand on the switch.” — Ante Projić

For Ante Projić, the value comes down to one distinction: “It’s the difference between running a server and running a publication.”

That distinction is only expected to sharpen. More of the operations loop is moving into conversation, an assistant watching monitoring, noticing anomalies, proposing fixes, with a human approving each one before it reaches production.

– The Ajme team, styled in coordinated blazers, gathers in the portal’s Split office. (Source. ajme.hr)

Ante Projić described the policy already in place: “We already work that way on Ajme by policy: the AI can read anything, but production changes need explicit human approval per item.” For a portal like Ajme, that division of labor means room to keep growing, richer archives, more automation, without growing the team.

Manage Your Servers With AI Assistants!

Try Cloudways MCP and manage servers, traffic, and deployments through AI assistants like Claude, all without leaving the conversation.

Try Cloudways MCP