Next week at DrupalCon Rotterdam, Pantheon will demo the “outside-in” path for Drupal AI working end to end. You’ll see an agent operating within Google’s enterprise AI platform handling the production work of publishing content created in Google Docs, Sheets and other Workspace assets.
This is the culmination of open-source contributions made via our AI Initiative sponsorship, plus our close relationship with Google Cloud Platform and what they’ve made possible via their Gemini Enterprise Agent Platform and SDK.
Get the Latest in Drupal!
Pantheon Drupal Agent displayed in the Gemini Enterprise interface.
Pantheon’s AI Initiative Sponsorship: MCP Server + Tools
Pantheon is a Gold sponsor of the Drupal AI Initiative, with a focus on making Drupal something that AI agents can use productively. This is what people sometimes call the “outside-in” approach, as opposed to the “inside-out” approach of embedding AI features like content tagging or translation within Drupal itself.
We engaged OMedia to do our sponsored contribution, as they were the original creators of the MCP Server module, and when I spoke with Giorgi Jibladze, their CTO, it was really clear we were aligned in our vision.
Giorgi Jibladze
Giorgi Jibladze
“AI powered features are cool, but we can see what big organizations really need to fit Drupal into a wider AI ecosystem. That’s what we’re most excited to work on, and it’s been great partnering with Pantheon to deliver this through the AI Initiative.”
–Giorgi Jibladze, CTO, OMedia
Our work, along with the contributions of many others, will soon bring the MCP Server and Tools ecosystem to a stable release, including a recipe and Composer packages that allow any modern Drupal site to easily start working with agents for site building, configuration and publishing.
Gemini Enterprise Agent Platform: Aligned With Drupal’s AI Principles
This year, our partners at Google Cloud Platform announced their enterprise AI solution: Gemini Enterprise. It builds on top of their industry-leading productivity suite, has a robust Agent Development Kit with freedom of choice for AI models, including open source, open weight and models from other frontier labs, and includes ironclad guarantees about data residency and privacy.
Naming things is famously “one of the two hard problems” in software engineering, along with cache invalidation and off-by-one errors, and people often assume that Gemini Enterprise is hard-locked to the Gemini LLM family, but that’s not the case. Google actually has what I think to be the most open and innovative approach relative to the other public clouds and frontier labs, one which aligns very well with the Drupal AI Principles.
For people who want freedom and privacy for their AI but don’t want the overhead and hassle of managing their own infrastructure, or who have fluctuating workloads that might be hard to predict, GCP is a great option. You can leverage their compute in any of 10 European national boundaries, using whatever model you like, be it Kimi or Claude, with full transparency into costs and clear guarantees around data and interaction privacy.
It was a natural fit to bring our work with the MCP Server and tools ecosystem to life by creating a Drupal Agent for Gemini Enterprise. Their Agent Development Kit is a real pleasure to work with, and we were able to get pretty far pretty quickly.
The Demo
Join us to see a new model for website administration: instead of asking every staff member to master the Drupal learning curve, teams can work through an interface they already use, connected to all the enterprise data they need, while approved agents handle complex workflows outside the CMS.
Editor's note: This first-person article was supplied to The DropTimes by Josh Koenig of Pantheon ahead of DrupalCon Rotterdam 2026. It has been edited for clarity, consistency and formatting while preserving the author's perspective and substantive claims. The DropTimes independently checked event and Drupal project details where possible. Google Cloud documentation confirms EU jurisdictional data-residency options and states that, for jurisdictional endpoints, “ML processing stays within that specific geographical region.” It also documents support for Google, partner and open models, with availability and residency conditions varying by model and endpoint. The DropTimes could not independently match the author's specific reference to “10 European national boundaries” in the public documentation available at the time of publication, so that wording has been retained as the author's account.
Image Attribution Disclaimer: At The Drop Times (TDT), we are committed to properly crediting photographers whose images appear in our content. Many of the images we use come from event organizers, interviewees, or publicly shared galleries under CC BY-SA licenses. However, some images may come from personal collections where metadata is lost, making proper attribution challenging.
Our purpose in using these images is to highlight Drupal, its events, and its contributors—not for commercial gain. If you recognize an image on our platform that is uncredited or incorrectly attributed, we encourage you to reach out to us at #thedroptimes channel on Drupal Slack.
We value the work of visual storytellers and appreciate your help in ensuring fair attribution. Thank you for supporting open-source collaboration!
Note: The vision of this web portal is to help promote news and stories around the Drupal community and promote and celebrate the people and organizations in the community. We strive to create and distribute our content based on these content policy. If you see any omission/variation on this please reach out to us at #thedroptimes channel on Drupal Slack and we will try to address the issue as best we can.
Josh KoenigBuilding the internets for a couple decades.
