Traditional CMS interfaces require editors to understand content types, fields, Views, workflows, and administrative navigation before they can complete a task. What if Drupal could instead assemble a focused interface based on what the editor wants to accomplish?
In this session, we will demonstrate an open-source Drupal module that connects Drupal AI providers with Google’s A2UI protocol. An editor can describe an outcome—such as reviewing launch blockers, planning a campaign, creating an article, or analyzing content—and Drupal generates a task-specific interactive interface using a trusted component catalog.
The module works through Drupal AI’s provider abstraction, allowing site owners to use OpenAI or another compatible provider without hardcoding a particular model. Generated A2UI messages are validated before being rendered using the official A2UI web-core and Lit renderer.
We will explore:
- How Drupal AI converts editor intent into structured A2UI messages.
- How A2UI can generate forms, review queues, dashboards, and editorial workflows.
- Why models should select trusted components instead of generating arbitrary HTML or JavaScript.
- How Drupal permissions, CSRF protection, action allowlists, and event subscribers create a safe execution boundary.
- How custom Drupal components could expose concepts such as moderation states, revisions, media, translations, and publishing schedules.
- Lessons learned while integrating an emerging UI protocol with Drupal.
The session includes a live Drupal demonstration using an actual AI provider, an inspection of the generated A2UI message stream, and a look at the module architecture and its open-source roadmap.