Instead of relying only on static information stored in the knowledge base, the Agent can call a function when it needs real-time data, account-specific information, or an external action.

Available in the PREMIUM plan and above.

With function calling, you can have your AI execute third-party APIs or custom logic using JavaScript code or HTTP Request to generate a dynamic answer.

When to use Function

The easiest way to decide whether to use a knowledge base or a function is to ask whether the information changes or requires an action.

Use the knowledge base for	Use a function for
Company information	Customer-specific information
Product descriptions	Order status
FAQs	Booking availability
Return and refund policies	Appointment availability
Shipping policies	Customer account information
Product documentation	Ticket creation
General instructions	Creating or updating records
Other relatively static information	Any live or transactional data

For example, your return policy can be stored in the knowledge base because it generally remains the same for multiple customers. However, if a customer asks “Where is my order?”, the AI Agent needs to retrieve the latest order information from your system using a function.

Do not use the knowledge base as a storage mechanism for information that changes frequently or is unique to each customer. Use a function to retrieve live information from the relevant system

Creating a function

Open Function Call

Open the canvas and add the AI Agent block.
Click AI Agent to open its configuration settings.
Select the required AI Agent from the dropdown.
Go to AI Agent configuration panel by clicking Pencil icon.

Create a function

Click Add Function>> Create FunctionGive the function a clear and descriptive name so that it is easy to identify when using it with an AI Agent.

Add the function description

Enter a description explaining:

What the function does
When the AI Agent should use it
What type of user request should trigger it
What information the function requires

The description is one of the most important parts of a function because the AI Agent uses it to determine whether and when to call the function

Configure the function handler

Choose how the function should perform the required operation.WotNot supports:

JavaScript — for custom logic and calculations
HTTP Request / API — for communicating with external applications and APIs

Save and Run Function

Once the function configuration is complete, click Save & Run function.Your function can now be connected to an AI Agent or used from supported AI blocks.

You can even create functions from Set AI and Answer AI blocks in Bot Builder screen also.

Function Components

Prompt
Function Handler

A brief description that helps the AI understand what this function does and when to invoke this function. It is important as you may create multiple functions, and the prompt helps the AI choose the right one.For example: “when the user requests the status of their order”

Function Types

Javascript
API/HTTP Request

Use JavaScript when the function needs custom logic that can be performed through code.JavaScript can be useful for:

Calculations
Formatting values
Modifying strings
Processing variables
Applying custom logic
Setting conversation or contact variables

WotNot supports running JavaScript on the Client or Server, and JavaScript can also be executed asynchronously when required.Example: Calculate a totalFor example, a function can calculate the total price based on a quantity and unit price:

const total = quantity * unitPrice;

The resulting value can then be used by the AI Agent when generating its response.Example: Modify a stringJavaScript can also be used to manipulate text values:

const customerName = name.trim();

This can be useful when you need to clean or transform information before using it elsewhere in the flow.

Use an HTTP Request function when the AI Agent needs to communicate with an external application or retrieve live information from a REST API.WotNot supports GET, POST, and PUT requests.ExampleSuppose your order management system provides an endpoint for checking order status.Request:

GET https://api.example.com/orders/{{order_id}}
Authorization: Bearer YOUR_API_TOKEN
Content-Type: application/json

Response:

{
  "order_id": "12345",
  "status": "Shipped",
  "estimated_delivery": "2026-08-15"
}

The function can then return the relevant information to the AI Agent, allowing it to respond:

Your order #12345 has been shipped and is expected to arrive on August 15.

The exact request format, authentication method, and response structure depend on the external API you are connecting to.

LIMITATIONS:

JavaScript is best suited for custom logic and data processing. If you need to retrieve live information from another application, use an HTTP Request/API function instead.

Testing a function

Test whether the AI Agent calls the function

Once the function works independently, test it through an actual conversation.

Connect the function to the AI Agent

Add the function to the relevant AI Agent configuration.

Test a matching user request

Ask a question that should clearly trigger the function.For example: “Where is my order #12345?”

Check whether the function was called

Review the conversation or function logs to confirm that the AI Agent selected the function.

Check the returned information

Verify that the AI Agent received the function result and used it correctly in its response.
By clicking on Setting icon , you can also change the configurations.

You can also check the logs by clicking View Logs

Test non-matching questions

Ask questions that should not trigger the function.This confirms that the function description is specific enough and that the AI Agent does not call the function unnecessarily.

If the function works correctly when you run it directly, but the AI Agent never calls it during a conversation, the first thing to review is the function description. The problem is likely with how the function’s purpose and trigger conditions are described, not with the function code.

Troubleshooting

The function is never called

Check the function description first.The AI Agent uses the description to determine when the function should be used. If the description is too vague, the AI Agent may not recognize that the function is relevant.Make the description more specific by explaining:

What the function does
When it should be used
What type of questions should trigger it
What information it requires

Also verify that the function is connected to the AI Agent.

The function is called at the wrong time

The function description may be too broad.For example:

“Use this function to get customer information.”

This could cause the function to be selected for many unrelated questions.Instead, make the trigger more specific:

“Use this function only when the customer asks for account-specific information such as their current subscription, account status, or plan details.”

Also clarify when the function **should not** be used.

The function is called with missing or incorrect parameters

Check whether the function clearly defines the information it needs.For example, if an order-status function requires an order ID, the AI Agent should know that the order ID is required before calling the function.Your function description can explicitly state:

“The order ID is required to retrieve the order status. If the customer has not provided an order ID, ask for it before calling this function.”

Also verify that the variable or parameter is mapped correctly.

The API returns an error

First, test the API independently using the function’s Test the API option.Check:

Endpoint URL
HTTP method
Headers
Authentication
Request body
Variable values
Required parameters

Also review the API’s response to identify the specific error returned by the external system.If the API works outside WotNot but fails in the function, compare the successful request with the configuration in WotNot.

The function returns data, but the AI Agent ignores it

If the function executes successfully but the AI Agent does not use the returned information, check the function’s output.Make sure the response contains the information the AI Agent needs and that the returned fields are mapped correctly.You should also review the AI Agent prompt. The prompt should instruct the agent to use the function result when answering the user’s question.For example:

“When you call the order-status function, use the returned order status and estimated delivery date when responding to the customer. Do not invent values that are not present in the function response.”

Authentication is failing

If the API requires authentication, verify that the required authentication details are configured correctly.Check:

Authorization header
Bearer token or API key
Header name and value
Token validity
Required Content-Type header
API permissions

For Bearer authentication, make sure the value follows the format required by the API.If the token has expired or been revoked, genera

Getting started

Build

Operate

Deploy

Reports

Account

Functions

When to use Function

Creating a function

Function Components

Function Types

LIMITATIONS:

Testing a function

Troubleshooting

​When to use Function

​Creating a function

​Function Components

​Function Types

LIMITATIONS:

​Testing a function

​Troubleshooting

When to use Function

Creating a function

Function Components

Function Types

Testing a function

Troubleshooting