This was my third Codegarden, but this year felt different, and not just because of the change of location but the vibe. After years in Odense, Codegarden returned to Copenhagen, where it all began back in 2005 with just 23 attendees. There’s something fitting about coming home, and it’s a feeling I know well. I’ve recently made a return of my own, moving back to Wales after years in Kent.
Attendance was at a record high, with 700 people in the room and hundreds more joining online from across the world. AI was everywhere, not as a buzzword being cautiously introduced, but as something the community has genuinely embraced and started building with.
Here are my highlights from day one.
Umbraco keynote
The keynote was split into two parts. Emma Burstow and Mats Persson opened with community news before Filip Bech-Larsen took the stage for the product announcements.
A series of AI hackathons held around the world, from Oslo to Manchester to London, have seen developers building packages in hours rather than days. Lotte Pitcher showcased some highlights, including a package that AI-extends images beyond their original boundaries and one that suggests redirects when content is trashed.
Filip’s technical keynote was packed. Three announcements stood out for me:
Elements is a new fourth native content type alongside documents, members and media. Non-routable, reusable content lives in a central Library section. Change it once and it updates everywhere it’s referenced instantly. Landing in Umbraco 18.
Umbraco Automate is a built-in low-code workflow automation engine. Define triggers (a page saved, media uploaded) and chain them to actions including AI agents, webhooks, and Slack messages. A “human in the loop” approval step lets editors review before actions complete. Free, open source, beta from conference day with general release in around four weeks.
Remote MCP moves the Model Context Protocol server to a fully hosted model on Umbraco Cloud. No local install, no tokens, just log in with your Umbraco ID. Free beta arriving this summer.
Running Umbraco in a changing world
Bjarke Berg, who leads Umbraco Cloud, covered the security realities of running Umbraco in 2026, and some of the numbers were eye-opening. Over the past 12 months, Cloudflare mitigated more than 5,000 DDoS attacks against Umbraco Cloud sites. The largest single attack peaked at over one million requests per second.
Bjarke announced Artefact Deployments, a second deployment model alongside Git, built for teams that already have CI pipelines in place. You build the artefact in your own pipeline, push it via a simple API, and Umbraco Cloud runs it. The key benefit is true rolling deployments with zero downtime. No cold starts, no dropped requests.
The longer-term direction is multi-cloud support via containers running on Linux. Artefact deployments are the first concrete step on that path.
Umbraco CMS in the age of AI
Bolette Kern, Product Manager for the CMS team, gave one of the more thoughtful talks of the day. Her central argument: Umbraco’s decision not to bake AI into the CMS core is now its most competitive position.
If every CMS ships the same out-of-the-box AI features, partners stop being experts and become installers of someone else’s AI strategy. Clients inherit model choices, provider lock-in, and costs they never agreed to. Umbraco AI acts as a governance and control layer instead, where you decide which models, which providers, and what the guardrails are, and everything built on top inherits those decisions automatically.
The guardrails demo was particularly compelling. Personally identifiable information is screened before prompts ever leave the CMS, and post-generation rules can flag misleading or unverifiable claims in generated content. All of it is logged.
The CMS Editor MCP demo was also impressive. Connecting Claude to a live Umbraco Cloud project and a Google Drive folder, she asked it to build a new landing page, pulling structure from an existing page, content from a meeting transcript in Drive, and images from a Drive folder. The whole thing happened without once opening the back office.
AI as your sidekick: Real lessons, real results
Barbara Forbes, an Azure Architect who spent six months letting AI run as much of her life as possible, gave a talk that was refreshingly grounded.
Three things stuck with me. First: AI is not always the answer. Her Todoist priority automation works brilliantly, but she was quick to point out that it’s not really AI, it’s automation. Knowing the difference still matters.
Second: AI should amplify you, not replace you. Her most viewed LinkedIn post in six months, 45,000 impressions, was a photo of her new desk. People want to connect with people, and AI content that replaces the human ultimately loses that connection.
Third: onboard your AI like you onboard a person. Invest in custom instructions. Label outputs as fact, interpretation, or speculation. Treat it like a junior engineer who will sometimes think it knows better than you.
This talk particularly resonated with me. Over the last few months I’ve been developing my own second brain using Claude and Obsidian, an ongoing experiment that has become part of my daily life. I use it for life admin and planning, and for this trip it came into its own, helping me sort my itinerary and suggesting public transport options. Barbara’s point about AI amplifying rather than replacing you felt very familiar.
It’s Elemental
Kenn Jacobsen’s deep-dive into Elements was the most technically detailed session of the day, and his enthusiasm was infectious. He called Elements the biggest change to Umbraco since language variants, and as someone who works with Umbraco every day, that claim didn’t feel like an exaggeration.
The Library section is the key idea: a central home for reusable structured content. Change an element once and it updates everywhere it’s referenced, instantly. And crucially, the rendering is completely unchanged. Your front-end code cannot tell the difference between a locally-sourced and a library-sourced element, and nor does it need to.
Kenn also mentioned that the Library section will grow over time. Content templates are an obvious candidate. Currently they live in the Settings section of the CMS, which means editors need Settings access just to use a feature that is fundamentally editorial. Moving them to the Library is long overdue, and something I’ve felt the friction of on real projects.
Umbraco Awards 2026
What better way to close out day one than with great food, great company, and the Umbraco Awards dinner? The food was fantastic, the atmosphere was electric, and the awards themselves were a brilliant reminder of just how much talent exists across this community.
As always, the quality of submissions across every category was seriously impressive. From enterprise-scale orchestration to beautifully crafted editorial experiences, the breadth of what people are building with Umbraco never fails to raise the bar.
For me though, there was a personal highlight tucked away in the results. Justin Neville took home the inaugural Umbraco AI Package Award for the AI Log Analyser, and Mike Masey was nominated in the Package Award category for the Rollback Previewer. Both make a huge effort for the Kent Umbraco community, and having recently moved on from Kent myself, it was genuinely brilliant to see two people I know recognised on the main stage in Copenhagen for their efforts.
Day two brought more sessions across AI, content workflows, and the wider ecosystem. You can read those highlights in part two of this write-up.
Thanks to Umbraco, all the speakers, and the community for another brilliant Codegarden.
#h5yr, you rock! 🤘