Bringing Drupal AI into your DNA: When you need to use code to create AI Agents & Tools!

Imagine asking your Drupal site not only to help you write content, but also to carry out tasks for you. It could create an event, update a configuration setting, organise information or even pass work to another specialised AI assistant. This is the direction Drupal is moving towards with AI agents.

In the fourth session of the Bringing Drupal AI into Your DNA series, we explored how these agents work and how developers can start building them in Drupal today.

The session began with a simple explanation. An AI agent is different from a normal chatbot. Instead of just answering questions, it can use tools to perform actions. For example, it can retrieve information from the site, create content, modify configuration or connect to external services.

Because of this, agents often work together. One agent might receive the initial request and decide what needs to be done. It can then pass parts of the task to other agents that specialise in specific actions.

We saw a few practical examples during the session. In one demonstration, an AI agent helped build a webform based on a simple visual layout. The agent first analysed the image to understand what fields were needed, and then used Drupal tools to create the form automatically.

Another example focused on managing content. An agent was asked to find certain articles and update their publication status. To do this, it searched the site for the relevant content and then used built-in Drupal actions to publish or unpublish the items.

Finally, we saw how developers can create their own tools for agents. In one demonstration, a user asked Drupal to create an event. The AI agent used one tool to create the event content in Drupal and another tool to add the same event to Google Calendar. From the user’s perspective, this happened through a simple conversation. Behind the scenes, the agent coordinated several steps to complete the task.

An important takeaway from the session was that good results depend on clear instructions. Developers need to carefully define what tools do and how the agent should use them. Well-written prompts and clear structure help the AI understand the task and reduce mistakes.

The session also looked at what comes next. The Drupal community is developing a new Tools API that should make these integrations more reliable and easier to reuse. The idea is that tools will not only work with AI agents but could also be used in other Drupal systems.

Security was another key topic. AI tools can be powerful, so they need to be carefully controlled. Future improvements aim to make it easier to define permissions and limits so that AI agents only perform actions they are allowed to do.

Finally, we saw early ideas for more visual ways to create agents. Instead of configuring everything through code, future interfaces may allow users to connect tools and workflows more easily.

The main message from the session was clear. AI in Drupal is moving beyond simple text generation. It is gradually becoming a system that can assist with real tasks across a website.

This work is still evolving, but the direction is promising. With the right safeguards and community collaboration, Drupal can offer powerful AI capabilities while keeping organisations fully in control.

If you are interested in knowing more, you can rewatch the session here.