The Functions section provides an interface for creating API integrations with your service and linking functions to agents.

A single function can be connected to multiple agents within the same project. This allows you to reuse functionality across different agents without duplicating configuration.

Functions can be enabled or disabled at any time, giving you full control over which capabilities are available to agents.

Function creation window

In the function creation window, you select one or more agents that will be granted access to the function.

Next, configure the API endpoint. Choose the request method (GET, POST, PUT, PATCH, or DELETE) and enter the endpoint URL. Depending on the selected method, you will need to define the required parameters such as path parameters, query parameters, or request body.

You must follow the exact structure shown in the examples inside the input form. Incorrect formatting may prevent the function from working properly.

You will also be asked to answer two important questions:

When this endpoint should be used

What this endpoint returns

Take time to provide clear and descriptive answers. These descriptions directly affect how accurately the AI understands and uses the function. Use the examples available under the “?” icon in each field as guidance.

In the final section, select the authentication type and enter the required credentials.

Test chat

The Test chat section allows you to test connected functions in real time by sending questions to the AI.

On the right side of the screen, you will see debug information that displays the requests being made, the tools that were used, and the arguments passed when calling your endpoint. This helps you verify that the function is working correctly and understand how the AI interacts with your API.