AI Decoupled Reaches Alpha With HTTP Endpoints for Drupal AI

The AI Decoupled module reached its first alpha release on 11 August 2026, exposing Drupal-configured AI processors through HTTP endpoints for external applications. Drupal architect Vincenzo Gambino announced the release in a LinkedIn post, describing support for AI agents, assistants, and direct model access. Drupal.org lists version 1.0.0-alpha1 for Drupal ^10.5 || ^11.2.

For decoupled projects, the module provides a route to Drupal-managed AI without requiring the interface to use a Drupal-rendered chat page. JavaScript single-page applications, mobile apps, static front ends, and server-to-server clients can call configured endpoints and receive output for rendering outside Drupal. The approach keeps AI processing and endpoint configuration in Drupal while allowing the interface to be built elsewhere.

Each endpoint can use an AI agent, assistant, or direct model executor. Administrators can restrict endpoints by role, configure per-endpoint rate limits, and enable server-sent event streaming where the selected processor supports it. Requests must pass a site-wide permission check, the endpoint's role restriction, and the selected AI processor's own access check. State-changing routes use credential-based authentication rather than requiring clients to provide CSRF tokens, and the module returns raw model text for the external client to render.

The alpha is not covered by Drupal's security advisory policy, and its release notes state that the API and configuration schema may change before version 1.0.0. Vincenzo credited Marcus Johansson, who manages Drupal's AI module, with providing direction and guidelines during development. Developers can find release details, documentation, requirements, and contribution information on the AI Decoupled project page on Drupal.org.