Umbraco launches new data orchestration solution – Umbraco Compose

Umbraco has launched its new data orchestration platform: Umbraco Compose. Umbraco says the platform will ease the management of data flows in complex enterprise projects. The solution is expected to save enterprises hundreds of development hours and struggles. According to Filip Bech-Larsen, CTO at Umbraco, “Compose is a data orchestration platform. It takes composable from something that is a buzzword into something that is reality. It makes it scalable but also maintainable.”

Meet Umbraco Compose

Bech-Larsen suggests it is tempting to describe Composable Architecture as a simple plug-and-play game. Systems that hold product, pricing, marketing, and image data are plugged in at one end, and frontends, websites, apps, screens, etc, are at the other end. However, this simple description underestimates one of the trickiest and time-consuming aspects of the setup – managing the “in between”.

Umbraco says its Compose is the “in between” data orchestration layer that ensures a clean and streamlined architecture. It includes a managed ingestion API and a headless delivery API using standardised GraphQL queries. This means that, regardless of the complexity of an organisation’s tech stack or the number of frontends needed to connect, Umbraco Compose will handle it. At the same time, keeping everything tidy and clear, as all data is managed in a consistent and structured way.

Clean Composable Architecture

Umbraco’s Data Orchestration Platform is based on composable technology. A modular design approach that breaks software and IT systems into independent, interchangeable building blocks, like Lego pieces. These components, connected by APIs can be assembled, reassembled, or replaced to create customized, flexible, and scalable digital solutions. It enables businesses to move away from rigid, “monolithic” systems to agile ones, allowing for faster updates and technology changes. Composable DXP, Composable CMS, MACH – words and concepts from just a few years ago that promised something powerful: freedom. Freedom from vendor lock-in, and the freedom to only select the systems and services you need to succeed.

Umbraco.ai: Foundation Package + Features

A base AI package that:

• Lets managers configure connections to different LLM providers (OpenAI, Gemini, Anthropic, Amazon Bedrock, Azure/Microsoft stack, etc.).
• Manages profiles (e.g., image recognition, text review) with model choice and parameters (temperature, tokens, etc.).
• Defines contexts (personas & tone of voice), e.g. an “Umbraco marketeer” persona using Umbraco’s tone of voice.
• Includes analytics, logs, and governance for debugging and control.

Feature packages on top of Umbraco.ai. AI is delivered through feature packages:

• Some built by Umbraco HQ
• Partners
• The Umbraco community

Examples launched / shown:

• Prompts package – store reusable prompts and attach them to specific fields.
• Agents / co-pilot package (preview) – an AI co-pilot inside the Umbraco back-office.

AI Inside the Umbraco Backoffice (Copilot & Prompts). Copilot capabilities (inside Umbraco): – Lives in the content editing experience.

It can:

• Generate a summary for a profile page and insert it into a subtitle field (with explicit user approval).
• Find and insert the correct author image from the media library based on tags and metadata.
• Always keep a human in the loop (you must approve actions before they’re applied).
• Browse the web to:
  • Fetch fun facts or information about a person (e.g., from the Umbraco team page).
  • Analyse competitor pages or other external content.
  • Help with things like reviewing legal docs or competitor positioning (as examples mentioned).

Prepared prompts for specific fields:

• Managers can attach one-click prompts to any field. Demo examples:
  • Alternative text (alt text):
  • A button next to an alt text field that generates alt text from the image.
  • You can regenerate until you like the result, then insert it.
• Meta description / SEO text:
  • A prompt that writes a meta description based on the existing page content.

• This pattern can be used on any field where editors repeatedly do similar tasks.

Key idea:
Anywhere in the CMS where editors do repetitive content work, you can wire up a prepared AI prompt to save time and keep consistency.

Agent Resources, MCP, and Skills

The Agent resources layer:

• On top of Umbraco CMS, they’ve created an “agent resources” layer:
• Works well with MCP servers so users can interact with Umbraco via agents in tools like ChatGPT, Claude, Gemini, or others.
• Supports skills so coding agents know how to build things correctly on Umbraco.

MCP & skills roadmap:

• Developer MCP:
• Already released for v16 and v17 and available via the marketplace.
• Makes it easier for developers to connect AI agents to Umbraco.
• Editor MCP:
• Coming next: a narrower scope MCP safe enough for editors.

Skills:

• First set: back-office extension skills so coding agents can create back-office extensions for the new back-office.
• Then more general implementation skills for broader Umbraco development.
• All hosted in an open source repo on GitHub, so you can try them now and provide feedback.

Open Source and Provider Choice

• Everything shown (foundation, prompts package, agent framework pieces) is open source.
• From day one, supported providers include:
• OpenAI
• Google Gemini
• Anthropic
• Amazon Bedrock
• Azure / Microsoft “foundry” stack
• Built on a provider pattern, so community and HQ can add more providers to the marketplace over time.

Business Model for AI

• Umbraco explicitly said they do not plan to monetise AI usage directly (no per token or per use AI charges from Umbraco’s side).
• Their AI strategy is to:
  • Grow adoption of Umbraco.
  • Emphasise credibility and flexibility rather than locking you into specific AI services.

Community AI Initiatives – Launching:

• An AI Community Team
• An AI Community Advisory Board
• New AI category in the Umbraco Package Awards 2026 to recognise innovative AI packages.

Encouragement to:

• • Build AI packages on top of Umbraco.ai and the feature packages.
  • Join hackathons (e.g., at Umbraco Spark in Bristol) to work on AI related packages.

Speaking at Umbraco’s Winter Keynote meeting on 3 February, Bech-Larsen described Umbraco as an AI Platform. He added,  “At Umbraco, we consider ourselves a platform for AI… It’s built for choice, it’s longevity, and it’s for real world use.”

Enterprise Times: What this means for businesses

Research by the MACH Alliance suggests a composable technology approach continues to gain adoption and generate a strong return on investment. Its 2025 Global Annual Research report found that 9 in 10 organisations that have implemented some composable architecture say it has met or exceeded ROI expectations. This is easy to understand.

A composable architecture is supposed to give businesses freedom, flexibility, scalability, and fast adaptability. This is increasingly becoming a necessity for many enterprise businesses. However, it must be managed and applied correctly to be successful in the real world. Some enterprises have found composable architecture to be challenging. It became painful upkeeping with custom endpoints. As Bech-Larsen suggests complex data management gave Composable DXP and MACH a bad rep. And with good reason, as the “freedom” suddenly came with an unexpected price tag from hours of debugging data flows and overloaded – slow – setups. Umbraco’s new solution is expected to simplify data management for enterprise projects and is set to save companies significant development time and effort.

The post Umbraco launches new data orchestration solution – Umbraco Compose appeared first on Enterprise Times.