Umbraco + AI: What exists today and what is on the way?
News
7 min
To say that AI development is progressing at a frightening pace may not be sticking your neck out directly. New models are coming almost every day, and the development does not seem to be slowing down anytime soon. But how can one implement and benefit from it in Umbraco? We will clarify that in this post!
When we wrote about Umbraco 17 (blog post 1, blog post 2) and its release last year, we mentioned Umbraco MCP as one of the major new features. Since then, quite a lot has happened, and Umbraco now has a much broader and clearer strategy for how AI should become part of the platform.
Umbraco divides this into two different areas:
- AI in Umbraco is about how we can use AI features directly in the editor experience and on the website. Everything from generating alt text to semantic content search.
- Umbraco in AI is the other way around. It’s about making Umbraco accessible to the AI tools we already use, such as ChatGPT, Claude, GitHub Copilot and other AI tools.
Let’s dig a little deeper!
AI in Umbraco
Umbraco’s philosophy around AI is that the CMS and the AI layer should remain separate. That’s why Umbraco has created something called Umbraco AI. It isn’t a ready-made AI feature in itself, but rather a shared layer that other AI features can build on.
A major advantage of Umbraco AI is that the solution isn’t tied to a specific model or AI provider. You can use models from OpenAI, Anthropic and Google, for example, and switch providers without having to rebuild the rest of the AI solution. And the best part? Umbraco AI itself is free and open source! The main cost instead comes from the API calls made to the models.
The basic idea is that the CMS itself should remain stable while the AI layer on top can evolve quickly. Pretty smart, considering an AI model can feel old and outdated just weeks after its release. To standardize how AI is used within Umbraco, functionality has been built around this, including Profiles, Context and Guardrails.
With Context, you can define your company’s tone of voice, writing guidelines, target audiences and other instructions that are useful for the AI to know. Profiles can then be created for different purposes. Your privacy policy, for example, probably shouldn’t use the same playful tone as the rest of the website, while your “Work with us” page probably shouldn’t sound overly legal and formal.
With Governance and Guardrails, you can set clear rules for the information sent both to and from the AI. For example, you could prevent the AI from using words such as WordPress, Sitevision and Optimizely on a page about Umbraco. You can also make sure certain sensitive information is blocked so that it isn’t sent to AI models at all.
So, what has been built on Umbraco AI?
Umbraco AI Prompt
With Umbraco AI Prompt, you can add predefined actions at field level. Let’s say you have a summary field on a content page. You can add a button next to the field that automatically generates a summary based on the text in the main content field.
Umbraco Copilot
Wish you had a little AI chat window on the side to help you while working with content in the Umbraco backoffice? That’s where Umbraco Copilot comes in. And the best part? It knows which page you’re on and understands the content and structure of that page. Using its other tools, it can also find and work with content from other parts of the website.
It can even make changes directly to the content on the page the editor is working on. For example: “Can you shorten the introduction and make it a little more sales-focused?” or “Can you fill in the SEO title and SEO description based on the content on this page?”

Umbraco Copilot Workspace
The next step is what’s called “Copilot Workspace”.
Instead of having the chat assistant sitting in the corner, it gets its own dedicated section in the backoffice. The focus here is more on larger tasks and working across multiple pages. It includes features such as chat history and Projects, where you can give the AI additional context and instructions.
At the time of writing, Copilot Workspace is available as a Release Candidate, meaning the feature is close to a full release but is still being tested.

Umbraco Knowledge Sets
Knowledge Sets let you give the AI additional knowledge and instructions it can use while working. This could, for example, include documentation and instructions related to an installed package. That way, the AI doesn’t have to guess how different parts of the solution work.
Umbraco AI Search
A traditional search is good at finding exactly what you type, but not much beyond that. With AI-powered semantic search, users can get relevant results based on the meaning behind their search, even when the exact same words aren’t used in the content. This means synonyms and similar phrasing don’t cause the same problems as they do with traditional text search.
That’s why Umbraco has developed Umbraco.AI.Search, where content is indexed for semantic search. This can improve the user experience and make it easier for visitors to find the content they’re looking for.
Umbraco.AI.Search is currently in beta, which means it’s available to test but is still under active development.
Umbraco Automate + AI
Another perfect use case for AI is Umbraco Automate. Automate is a package developed by Umbraco for building automated workflows using triggers and actions. And AI can be added to those workflows too.
For example, why not automatically let AI generate alt text when a new image is uploaded to the Umbraco media library? Or ask AI to summarize what has changed when an editor publishes new content and then send that summary as a Slack message? Or automatically create a draft English version whenever new Swedish content is published? The possibilities are endless!
The AI integration with Automate is still in alpha and therefore at an early stage of development.

Umbraco Engage.AI
Umbraco’s marketing package, Umbraco Engage, has also gained the ability to connect with Umbraco AI. With the help of Umbraco Copilot, you can now get assistance identifying which campaign performs best, which pages perform best or how different audiences behave.
Umbraco Engage.AI is currently in public beta, meaning it’s available for public testing but is still being developed and may change.
Build your own
In addition to the existing packages, you can of course build your own solutions on top of Umbraco AI. It’s there as a foundation to build on. So the only limit is your imagination!
Umbraco in AI
Now it’s time to turn things around. Instead of bringing AI into Umbraco, we’ll focus on the opposite: how can we make Umbraco accessible to AI?
Open Source
Another advantage when it comes to AI-assisted development is that Umbraco is open source. Since the source code is open, AI agents can access it and investigate how the CMS actually works instead of relying solely on documentation.
MCP
MCP stands for Model Context Protocol, a term that’s appearing more and more when talking about AI agents. It’s an open standard for how AI agents can connect to different systems and use their tools and data in a structured way. And, of course, Umbraco is one of those systems.
With MCP, Claude, Codex or other AI clients that support MCP can communicate with and perform tasks directly in Umbraco instead of you having to do everything yourself. And there are several MCP servers available.
For developers, there’s the Umbraco CMS Developer MCP, which can be used when developing Umbraco websites. It can, for example:
- Create document types
- Create and update content
- Work with media
- Create Dictionary Items
- Manage data types
- Analyze the structure of an Umbraco installation
Umbraco Forms MCP
There’s also an MCP for Umbraco Forms. It allows an AI agent to work with forms, data sources and submitted form responses, among other things. Forms MCP can also be connected to CMS MCP, allowing an AI agent to work with both content and forms within the same workflow.
Umbraco Commerce MCP
Umbraco Commerce has its own MCP as well. Here, AI agents can work with orders, discounts and analytics data from Commerce, among other things. Commerce MCP is still a Proof of Concept, meaning it’s an early implementation primarily intended to demonstrate what’s possible.
For editors, there’s a customized version called CMS Editor MCP. Instead of focusing on the structure and development of an Umbraco website, it focuses on the content itself. For example, you could ask it to:
“Find all news articles from 2025 that are missing an introduction.” or “Create a draft for a new page based on this brief.”
And the best part? It will also be possible to connect an AI client directly to your Umbraco Cloud project and work with the environment through what’s called Remote MCP in Umbraco Cloud.
The roadmap also includes a Content Delivery MCP, which gives AI read-only access to published content and media through Umbraco’s Delivery API. This means AI tools can analyze and understand what is actually published on the website without being able to change the content.
To sum it up:
- Developer MCP: build and modify Umbraco.
- Editor MCP: work with content.
- Forms MCP: work with forms and submitted responses in Umbraco Forms.
- Commerce MCP: work with data and functionality in Umbraco Commerce.
- Content Delivery MCP: understand published content, read-only.
Umbraco CMS Developer MCP is available today. CMS Editor MCP and Hosted Remote MCP for Umbraco Cloud are currently under development, with Q3 2026 as the target release. Content Delivery MCP is also on the roadmap but currently has no announced release date.
Skills
MCP gives AI agents the tools they need to do things in Umbraco, but they also need to understand how those tools should be used. That’s where Skills come in.
Skills are structured instructions and best practices that help AI agents understand how Umbraco works and how to work with it effectively. Today, there are more than 60 Skills for developing the Umbraco backoffice, with more on the way for areas such as content modeling and general implementation.
AI from two directions
Perhaps the most interesting part isn’t any single AI feature, but the fact that Umbraco is approaching AI from both directions. On the one hand, AI is becoming a natural part of the CMS itself. On the other, Umbraco is becoming accessible to the AI agents and tools we already use.
As you can see, there’s a lot happening in Umbraco when it comes to AI, and new things are being introduced all the time. Exciting times ahead!