Major Release

The Long-Term Supported version of Umbraco is here

Umbraco 17 LTS release →
Book a discovery call
Book a discovery call
Umbraco and AI

Flexible Foundation for an AI Future

Enabling flexible, future-proof digital experience with AI on your terms.

Explore the CMS
Umbraco

Our AI Strategy

The strength of Umbraco has always been its ability to empower developers, editors, digital agencies, and enterprise clients with a flexible, open CMS platform built to support how you want to work. That same principle guides our AI vision: not a rigid, “one-size-fits-all” layer, but a set of powerful, optional tools you control.

We believe AI should work for you, not the other way around. Whether you're a developer extending the platform, an editor creating content, or a business optimizing digital experiences, AI with Umbraco is designed to enhance your workflows, not replace them.

We see AI as an opportunity to extend Umbraco’s flexibility, not restrict it. By focusing on standards, openness, and developer choice, we’re making sure AI fits into the workflows and solutions you want to build.

— Filip Bech-Larsen, CTO, Umbraco

Freedom to choose your AI stack

Umbraco doesn’t lock you to one LLM. Use OpenAI, Anthropic, Microsoft, or your own model. The choice is yours. It stays open and flexible so your AI stack can grow with your strategy.

Adoption on your terms

AI should fit your workflows - not force new ones. With Umbraco, you adopt and adapt AI at your pace, keeping full control of data, governance, and how deeply AI shapes your processes.

Open, collaborative innovation

Innovation thrives in the open. Locking your CMS to one vendor limits future choices. With Umbraco’s Marketplace, you can use AI add-ons, ready solutions, or build your own.

An Extensible Content Platform Built for AI

Umbraco provides a clean, structured content foundation, open APIs, and transparent architecture that AI tools can understand - and the extensibility to build the intelligent experience your teams need.

  • Developers get reliable, well-defined data that works with LLMs and frameworks like Semantic Kernel.

  • Editors benefit from consistent content patterns that enable context-aware assistance.

  • Businesses gain a solid, future-proof base for automation, search, personalisation, and modern content workflows.

Umbraco gives you a platform that’s easy for AI to work with — and easy for you to extend — so your digital experiences can evolve as the technology does.

In our experience, Umbraco’s flexible architecture, intuitive content model, and open APIs make it extremely easy to integrate AI, both on the client side and within our internal workflows.

— Beckie Dilworth, Chief Strategy Officer, BlueModus

The Era of Agents

AI Codegarden.jpg

We’re entering a new chapter: AI agents — systems that can reason, take action, and complete multi-step tasks.

Agents turn passive AI into active collaborators. For example:

  • “Create a new landing page, draft content, and add relevant media.”

  • “Audit all published pages for missing SEO and AEO metadata and propose fixes.”

Agent workflows break a user’s goal into multiple steps. The path can vary (the agent explores, retries, and adapts), but the outcome must pass clear success tests. You don’t need to specify every step—just the goal and the validation. The agent uses creativity to get there; your checks make the result reliable.

This era aligns perfectly with Umbraco’s strengths. The clear content architecture, predictable APIs, and structured editorial model give agents exactly what they need: stable, well-defined data and rules they can interact with safely.

And that’s why we created the Umbraco MCP Server: the bridge that lets agents read, write, and act in Umbraco using well-defined tools that reflect how Umbraco CMS works, safely within controlled boundaries.

A Spectrum of AI Experiences

For Developers: The Umbraco CMS Developer MCP

The Developer MCP Server mirrors the full Umbraco backoffice. Anything you can do in Umbraco, you can do through MCP — safely and locally.

Developers get:

  • Complete access to the CMS via the Management API

  • Local-first MCP server for full control

  • Compatibility with tools like Claude Code, Cursor, Windsurf

  • Flexible tooling and the ability to customize or extend the Umbraco MCP server

  • Tool collections to limit scope, improve performance, and reduce cost

Ideal for human-guided automation, rapid prototyping, scaffolding site structures, data migrations, and building your own AI-powered developer workflows.

Learn more

We first experimented with early versions to clean and restructure a legacy CMS we maintain—reorganising content types, adding labels, refining properties, and generally bringing long-standing content models under control. Its power and flexibility have made it a core part of our workflow.

More recently, we’ve used the MCP Server on a new Umbraco build, leveraging it to generate a clean base solution guided by specification-driven development using Claude Code and Cursor. It has also proven invaluable for content migration from external systems, something we’re actively using for a current client project.

— Oliver Picton, Head of Technical, Enjoy Digital

For Editors: The Hosted, Task-focused Editor MCP

Editors need focus and simplicity, not complexity.

That’s why the Editor MCP Server (coming soon) is built specifically around everyday editorial tasks — and fully hosted by Umbraco, so anyone can start using it in minutes.

With the Editor MCP, editors get:

  • No setup, no installation, no developer needed — just connect your site and start prompting.

  • Task-focused tools that handle real editorial work: rewriting, translating, summarizing, restructuring content, improving accessibility, refining SEO, generating metadata, analyzing tone or consistency, optimizing image alt text, detecting duplicated media, and more.

  • Natural-language workflows — editors describe the outcome, not the steps.

  • Safe, permission-aware actions that respect existing roles, approval workflows, and publishing rules.

This means editors can do things like:

  • “Rewrite this article to better match our company tone and shorten it by 20%.”

  • “Analyse this page and suggest improvements for clarity, accessibility, and SEO.”

  • “Translate this product section to French and adjust phrasing for a retail audience.”

  • “Scan this campaign folder and generate missing alt text and meta descriptions.”

Looking Ahead

The foundation is set, and the possibilities are vast. From the Developer MCP available today, to the upcoming Editor MCP, we’re creating a framework where AI can safely and powerfully interact with Umbraco across every layer.

Because the future of CMS isn’t just content management.

It’s content intelligence — open, extensible, and powered by Umbraco.

Explore the CMS

AI thumbnail.png

FAQ

Umbraco MCP (Model Context Protocol) is a way for AI agents and AI clients (like ChatGPT, Claude, Cursor, etc.) to interact safely and intelligently with Umbraco. It translates Umbraco’s APIs and content structure into tools an AI can understand, allowing automation, content creation, analysis, and developer workflows.

Yes. The Umbraco CMS Developer MCP Server is fully open source. You can inspect, extend, and contribute to the code, and run it locally for complete transparency and control.

You can get the Umbraco CMS Developer MCP Server here.

An AI agent is a system that can reason, take action, and complete multi-step goals, not just answer questions. Some agents are autonomous, handling entire workflows end to end. Others are reactive, performing actions step by step under direct control.

For example: “Create a new landing page, draft content, and add the appropriate media.”

Agents break this prompt into LLM steps, retry when needed, and deliver a validated result.

AI clients (like ChatGPT) already contain general-purpose agents.

When a user opens ChatGPT and connects it to Umbraco MCP, ChatGPT becomes the agent.

It automatically uses the MCP tools that Umbraco exposes. No coding needed.

Digital agencies can build custom agents.

For workflows like:

  • Automated QA

  • Structured content migration

  • Bulk rewriting and auditing

  • Design → build → test pipelines

These agents run against the same Umbraco MCP tool.

Organizations can build internal agents. For custom governance, content review, compliance, or editorial automation.

Umbraco MCP Server is designed for transparency, predictability, and human oversight. You stay in control because:

You decide what the AI can access. You enable only the MCP tools you want—this defines exactly which parts of Umbraco the AI is allowed to work with.

Umbraco permissions still apply. Role- and permission-aware actions ensure AI can only perform tasks the underlying user is allowed to do.

Full visibility during development. With the local Developer MCP, your AI client shows every step the agent takes, including the tools it calls and the actions it performs.

Editors get a safe, limited environment. The hosted Editor MCP restricts AI to guided, task-focused editorial actions—no schema changes, no surprises.

Everything is logged and reviewable. All MCP interactions are API calls, meaning every step the AI takes can be monitored, audited, and traced.

The AI never has more access than the user behind it.

You do, always. Umbraco does not claim ownership of your content, your prompts, or the data you process through MCP. Nothing about AI changes that.

Your content stays yours, your business rules remain yours, and your governance strategy stays yours, whether you use AI lightly, heavily, or not at all.

Only if you explicitly choose to share it.

Umbraco MCP Server is model-agnostic and works with any AI provider, including fully self-hosted or private models.

This means you decide where your data goes and which model processes it.

If you choose a cloud LLM (such as GPT or Claude), your data is governed by that provider’s enterprise privacy terms. Depending on your tier or contract, this can include strict non-training guarantees and region-specific data handling.

Yes. Umbraco MCP Server works with self-hosted or private LLMs, making it possible to keep all data inside your own infrastructure when needed. You can connect MCP to enterprise-grade platforms such as Azure AI Foundry, Amazon Bedrock, or Google Vertex AI, or to your own models running through tools like Ollama, vLLM, or a private Hugging Face deployment.

Using a private or on-premise LLM gives you full data isolation, meaning prompts and content never leave your environment. It also gives you complete control over model versions, updates, and fine-tuning so you can optimise the AI for your organisation’s terminology, content, and compliance needs. For teams in regulated sectors, from public institutions to healthcare and finance, this provides the confidence of strong governance with all the benefits of AI-assisted workflows.

Umbraco focuses on openness, choice, and user control, the foundations of ethical AI adoption.

Users choose their models, hosting, data boundaries, and governance rules. Umbraco avoids lock-in, forced pipelines, or opaque AI behavior.

Umbraco MCP is designed to help teams use AI responsibly, including reducing environmental impact. MCP’s tool collections let you control exactly which capabilities are active at any moment. These collections act as quick toggles: you can start with a lean setup, switch on additional collections as your workflow grows, and scale back to a minimal set when you’re done. Fewer active tools means less computation and lower energy use.

Because you can choose where and how the model runs, from efficient cloud models to greener regional datacenters to local processing, you’re free to optimise for sustainability as part of your AI strategy. The result is a practical, flexible way to keep your AI footprint as small as possible without compromising capability.

Yes. Umbraco MCP Server is explicitly built to enable multi-tool, end-to-end AI workflows.

Teams today are using MCP alongside tools such as:

  • Figma MCP → generate components, layouts, and structure directly from design files

  • Playwright MCP → test UIs the AI just created

  • Git MCP / GitHub MCP → commit changes or trigger workflows

This enables “design → build → test → publish” automation loops powered by AI.

Yes, and this is precisely where AI agents excel.

You can automate tasks like:

  • Building content structures from Figma and/or Storybook components

  • Drafting and refining content

  • Running accessibility checks

  • Updating metadata

  • Publishing structured content

  • Automate debugging and resolve issues from logs and health checks

  • Running automated tests against UI

Umbraco MCP Server can make these actions predictable, safe, and repeatable.

The Developer MCP is available today.

The Editor MCP launch is planned for Q2, 2026.

The Content Delivery MCP is planned for later in the roadmap, following the Editor MCP.

All AI and MCP features are tracked in the Umbraco Product Roadmap.

No. It’s completely model-agnostic. You choose the LLM and hosting setup that fits your strategy.

Yes, the difference is that, by interacting directly with the underlying APIs, agents don’t need to click through the interface step-by-step. This makes working with Umbraco faster, more efficient, and far more flexible, enabling new levels of creativity and automation that go beyond what the UI alone can offer.

Absolutely. The Umbraco MCP is a powerful foundation for building tools across modern agent ecosystems. Its fine-grained control and flexible tool collections make it a natural fit for Agent SDKs from OpenAI, Anthropic, CrewAI, Pydantic, and more. Whether you're orchestrating intelligent workflows or extending your AI capabilities, the Umbraco MCP gives you the precision and adaptability you need to build exceptional agent experiences.