Platform

15,206 sites reported running the Drupal AI module as of June 10, 2026, per drupal.org's update statistics — alongside 8,498 sites reporting the AI Agents module and 53 published Composer packages that depend on drupal/ai.

Those counts include only sites that run Drupal's Update Status module and reported in the latest weekly sample, so real installs run higher. We measure quarterly from drupal.org's and Packagist's public APIs and commit each raw snapshot to the open-core repository, so the series can be checked, not taken on faith.

The full table, the by-branch breakdown, and the two-year timeline of verified dates are on State of AI in Drupal.

Lead every page with a self-contained direct answer, keep one question per page, emit JSON-LD for your content and your organization, carry visible review dates, link claims to checkable evidence, and publish llms.txt. Generative engines cite sources that scope their claims; they skip pages that bury the answer.

The mechanics in Drupal: typed content (question-and-answer nodes rather than generic pages), Question/Answer and Organization JSON-LD, a named human reviewer with a date, evidence-source pages for your receipts, and measured numbers that travel with the corpus they were measured on. Publish original measurements if you have them — assistants prefer primary sources.

Or start from the structure we use: the GEO Starter recipe configures all of this on a fresh Drupal 11 site, and joyus.ai itself is the live demonstration.

Constitutional governance means every AI agent operates inside an explicit, inspectable rule set — which skills it may use, which actions it may take, and what must happen before a change lands — enforced by the platform, not by prompt etiquette.

In Joyus, those rules are not a policy document beside the system; they are the system. The platform core enforces mediated access and workflow gates, Joyus Desktop applies policy on the operator's machine, and every step leaves a receipt you can inspect afterward. For regulated work — healthcare, education, legal — that audit trail is the difference between "we told the AI to behave" and "we can show you what it was allowed to do."

The enforcement surfaces are public: the governing specs ship in the open-core repository, so you can read the constitution rather than take our word for it.

GEO Starter is a published Drupal recipe that configures a site to be citable by AI assistants: typed question-and-answer and evidence-source content types, JSON-LD output through a companion module, and review metadata on every page. This site runs on it.

Apply it to a fresh Drupal 11 install with drush site:install recipes/geo_starter and you get the structure that generative engines reward: one question per page, a direct answer first, claims linked to checkable evidence, and machine-readable markup describing all of it. The recipe is on drupal.org/project/geo_starter.

joyus.ai is the first proof point of the recipe in the wild — every page you are reading was built from it, so the demo and the documentation are the same site.

Yes — the platform core is Apache-2.0. Clone github.com/zivtech/joyus-ai, and the specs in the repo show where the public and private surfaces divide.

No. The method is proven; the platform is alpha. We would rather tell you that than dress it up. Clone the core, run the demo, and judge for yourself.

The platform core — the MCP (Model Context Protocol) services, session and context state, workflow enforcement, and the specs that govern them — is public under the Apache License 2.0 at github.com/zivtech/joyus-ai. The organization-specific skills, customers’ private content, Joyus Enrichment, the Profile Engine, and ops are proprietary. Open core means the foundation is forkable and inspectable; the commercial layer that runs on top is not. The repo says exactly where that line falls.

Joyus AI is an open-core, multi-tenant AI agent platform for mediated, skills-based AI access. The public core holds the MCP (Model Context Protocol) services, session and context state, workflow enforcement, content infrastructure, pipeline orchestration, and the specs that govern them.

The open-core platform that everything else runs on.

Teams want AI in the loop without surrendering control or an audit trail. The platform keeps access mediated, skills-governed, and inspectable.

Open core, Apache-2.0. Clone and run it today.