Our goal was to help you assess whether extending your declarative agent with an action via an API plugin is suitable for your scenario. We developed some criteria to guide your decision:

• The agent needs access to data beyond its foundational model.
• The external data is exposed via an API.
• The API uses a supported authentication mechanism.
• The agent must read real-time data.
• The agent must modify the data.
• The external data the agent needs to access is structured.

We applied these criteria to a fictional scenario of an auto repair shop needing an AI-powered assistant to query repair information stored in an external system. Extending a declarative agent with an API plugin meets all the requirements and allows room for future needs.

Using these criteria helps you evaluate whether using a declarative agent with an API plugin is appropriate for your scenario. The criteria allow you to assess the type of information the agent needs to access, how that information is exposed, and how the agent should interact with it. By using these criteria, you have enough information to choose the right approach for your scenario.