Mastering AI in Umbraco: The rise of contextual AI with MCP

Please note our content disclaimer in relation to blog posts.

AI has surged across the industry, bringing both breakthroughs and blind spots. It’s transforming how we build, manage, and think about websites, including those powered by Umbraco. The recent announcement of official MCP (Model Context Protocol) support from Umbraco HQ cranks this excitement up even further.

But let’s not get ahead of ourselves.

Let’s not trust the internet, it’s a sloppy place

Prior to the advent of artificial intelligence, the internet contained a substantial amount of questionable advice, outdated code samples, and improvised solutions that were often not reliable. Any developer who has ever trawled endless Stack Overflow threads knows exactly what we mean.

Of course, the internet isn’t all bad. It’s still a treasure trove of knowledge and brilliant ideas if you know where to look.

AI has not resolved this issue; in some cases, it may have amplified it. Large language models (LLMs) can incorporate unreliable content and occasionally present it with strong certainty. This can create a cycle of misinformation that poses challenges even for experienced developers.

This does not imply that AI is inherently flawed or unreliable. It emphasises the need to regularly assess suggestions carefully and highlights the importance of understanding the subject matter. These practices are not new; they have always been essential, even before the rise of AI, to ensure the reliability and accuracy of our decisions.

AI: Your assistant, not your boss

Here’s a golden rule. Never let AI solve problems you don’t understand yourself. That’s a fast track to a messy codebase, sleepless nights, and technical debt.

Think of AI as your supercharged research assistant or your digital rubber duck (yes, we’re big fans of talking to inanimate objects to debug problems). It can summarise documentation, suggest improvements, and handle the tedious tasks you’d rather avoid. But at the end of the day, you’re still the one steering the ship.

When using AI, it’s vital to provide clear prompts and proper context to guide it effectively. Without this oversight, outputs can be inaccurate or miss the bigger picture. Mastering the art of prompting AI is becoming as important as developing your “Google-fu” to dig up the right answers.

MCP: Finally, AI you can trust

This is where MCP steps in. As outlined in Umbraco HQ’s announcement and detailed by MCP creators Phil Whittaker and Matthew Wise in their blog, MCP gives AI a much-needed sense of direction.

Instead of relying on a general-purpose model trained on the entire internet (and all its bad habits), MCP lets you define trusted contexts like your Umbraco setup, content model, or business rules. You get answers that make sense for your project, with no more AI guesswork or suspicious shortcuts.

What does this mean for developers?

For developers, MCP is like having an extra team member who never takes a holiday or finishes the office biscuits. It offers deep insights and a better understanding of your entire Umbraco setup.

Many of these features are still theoretical, but MCP could potentially:

• Identify repeated CMS elements across your project
• Suggest consolidating document types for better maintainability
• Highlight potential compositions within your content model
• Tidy up image crop options for consistency, identifying common crop ratios
• Improve the editor experience by recommending property orders, groupings, or better naming and descriptions
• Validate redirects, check for conflicts and overlaps, and help prevent redirect loops
• Consolidate redirects into cleaner patterns, such as using regex

Although these concepts have yet to be extensively validated in practical applications, their potential is significant. An AI that thoroughly comprehends your site's architecture can facilitate the development of a streamlined and sustainable Umbraco setup. MCP is beginning to deliver on this promise by establishing a strong foundation for implementation within trusted AI models.

What about site owners and editors?

It’s not just developers who benefit. Site owners and editors also stand to gain a wealth of practical, actionable insights. With MCP-powered tools, teams can make better decisions, improve user experiences, and stop guessing about what might work.

Possible ways MCP can help include:

• Showing how visitors interact with your site
• Identifying weak pages or areas needing improvement
• Enabling data-driven decision-making based on real user behaviour
• Highlighting pages with high exit rates and suggesting fixes
• Recommending conversion funnel improvements
• Proposing content restructures based on user trends
• Providing richer insights when integrated with Umbraco Forms, Umbraco Commerce, and Umbraco Engage
• Revealing drop-off points and common form validation issues
• Suggesting corrections or restructuring forms using data from Umbraco Engage and Forms

Looking forward, editors could gain even more. MCP might:

• Reveal which content resonates best with different audiences
• Suggest opportunities for content repurposing
• Support ongoing SEO improvements
• Integrate with Umbraco Workflows to provide automated, first-pass content validation before human review
• Catch issues early, improving confidence and editorial flow
• Ensure everything aligns with brand guidelines and technical standards

As mentioned before, these possibilities are still theoretical, but they hint at a future where MCP makes AI not just a helpful tool, but a genuinely trusted partner in content creation and management.

The bottom line: AI needs you

AI offers various capabilities, but its use requires careful management. Integrating large language models with a context-driven approach such as MCP can enhance both processing speed and accuracy while supporting creative and controlled outputs.

At Cantarus, we’re buzzing about where this is going. We see massive potential for Umbraco sites and can’t wait to help our clients unlock smarter, more powerful, and more meaningful digital experiences.