Is Umbraco MCP Server Usable in Practice, or Is It Just a Phase?

Fifteen years in the Umbraco ecosystem have taught me a few things.
So naturally, when Umbraco MCP Server – the shiny new “connect your prompts to your CMS” automation invention – entered the scene, I asked myself the same question you probably did:

Is this actually useful, or is it just another tech phase — like NFTs, Google+, or that time people tried to use headless CMS for literally everything?

Let's explore.

Umbraco MCP: Connecting Prompts and CMS … Because Why Not?

If you haven’t heard the hype:
MCP Server essentially lets Umbraco talk to AI services in a structured way.

So basically, we can ditch the backoffice UI and just prompt our way to creating pages? To quote Queen: “It’s a kind of magic.”

Or is it?

But from a non-developer (yet tech-interested) perspective?
It’s kind of like owning a Formula 1 car but only being allowed to drive it to the grocery store.

How Non-Developers Could Actually Benefit

Finally, I can do things I’d normally need developers to handle.
Honestly, I’m already imagining my “one-man-army agency” where UX, design, copy, animations, integrations and CMS implementation can all be done by one person – me.

But then again … if something seems too good to be true, it probably is.

1. Content Automation (AKA: Make the CMS Do the Chores)

Okay, let’s dream for a moment.

Imagine telling Umbraco:

“Create a landing page based on this PDF, add SEO, fix the tone — and don’t embarrass me.”

But it didn’t – and it probably couldn’t.

After running some tests (with my dev-bestie Brian in the driver’s seat), it became obvious:

The more creative the task, the less useful MCP becomes.

We discovered that we can’t meaningfully influence how our prompts are interpreted by the Content Delivery API. Our theory?

The API documentation simply isn’t yet written with prompting in mind.

We received too many errors.
But it was honestly fascinating to watch the conversation between Claude and what Umbraco returned.

Because let’s be honest… automation is great until it isn’t.

The biggest issue with AI isn’t the technology — it’s the discipline of prompting. No one ever taught me how to prompt properly. And even if there were a manual, I probably wouldn’t read it. Just one of those things… maybe undiagnosed, maybe sporadic ADHD. For instance:

The Overconfident Prompt

Tell MCP:

“Generate a product description.”

It produces a philosophical essay on the metaphysical meaning of Bluetooth speakers.

The Infinite Loop of Doom

Set up chain-of-thought prompts.
The CMS starts generating content based on content it generated based on content it generated…

Congratulations — you’ve created a self-aware SEO spam monster.

Ok, to be fair, these examples are thought experiments.
But the real issue is this:

With MCP we become the co-driver — not the driver.

We can warn the driver that we’re switching from tarmac to gravel,
but how the driver handles the transition? Only time will tell.

Sure, we can keep adjusting our directions,
but ultimately, the one behind the wheel chooses the strategy.

The Dream

Let’s imagine a CMS where:

• You describe what you want.
• Prompts don’t have to be hyper-specific.
• Content updates itself intelligently.
• Workflows work with you, not against you.
• Minor changes don’t involve Jira tickets, 14 Slack messages and a 2-hour dev meeting with diagrams nobody understands.
• That’s the promise — or at least the hope — behind Umbraco MCP Server:

A world where CMS automation is actually fun, and maybe even useful.

Why Does Umbraco Even Bother?

Good question.

1. Staying relevant in the AI Era

If you don’t say “AI” at least six times per keynote in 2025, do you even exist?

2. Developers Were Getting Bored

And bored developers are dangerous.
They will build something – and this time it’s MCP.

3. Because “Headless” Isn’t Buzzword Enough Anymore

Needed something fresh for conference slides.

4. They Actually Believe in It

Honestly? Probably this.

Umbraco has always been good at making tech human-friendly.
MCP is simply the next step: Make automation less terrifying for normal people.

So … Is It Just a Phase?

Probably not.

MCP Server feels like the kind of thing that:

• Developers will obsess over (once they can properly control it),
• Content teams will eventually rely on,
• Non-developers will pretend to understand,
• And consultants will explain to customers for the next decade

It might be confusing, chaotic and capable of spectacular fails…

But it’s also a glimpse of the CMS future.

We’re still early.
Right now, MCP is best suited to simpler tasks like:

• “Which products perform best during campaigns?”
  (Engage + Commerce know that stuff.)
• “Check this page for SEO issues.”
• “Give me all unpublished pages.”
• etc.

But my real hope?
Enabling the “non-editors.”

Employees full of knowledge who never go near a CMS.

Our B2B clients often struggle to collect real customer insights.
Imagine:

A sales manager brain-dumps context about a client.
MCP turns it into a case study draft and notifies an editor.

Or:

A sales manager needs a big pitch.
They prompt a personalized landing page tailored to the prospect.

Either way - Long live MCP.

Maybe the possibilities exceed the needs.
Maybe I’m overly optimistic.

Or maybe I’m just being hopelessly optimistic.