Built-in and Custom Tools

​

Built-in Tools

​

AI Training

• The AI Agent will automatically look for messages left unanswered by team members in ongoing conversations
• Uses these unanswered messages to train and improve itself over time

​

Allow AI to Send Messages

• When enabled, the AI Agent is permitted to send messages directly to customers
• Disable this if you want the AI to only assist internally without responding to customers directly

​

Allow AI to Create Tickets

• Allows the AI to automatically create tickets based on the situation and conversation context
• Tickets are used for issues where features are not working as expected, and complaints or problems the customer is facing
• The AI should create a ticket when any of the following conditions are met:
  • The AI has not been given enough information to solve a complex problem
  • The customer explicitly asks to speak with a human agent
  • The customer is upset or frustrated

You should create a ticket when any of the following conditions are met:
- You haven't been given enough information to solve a complex problem
- The customer explicitly asks to speak with a human agent
- The customer is upset or frustrated

Important: The AI must never create a ticket if a ticket has already been raised in the same conversation on the same topic. Instead of creating a duplicate ticket, the AI should respond to the customer directly.

​

Allow AI to Create Private Notes

• Allows the AI to create private notes visible only to your internal team
• For minor questions or TODO items for the team, the AI can create a private note using the privateNoteToHuman field

If it is a minor question, you can create a private note by adding a
message to the privateNoteToHuman field.
If you need to create a TODO item, you can create a private note by
adding a message to the privateNoteToHuman field.

​

Custom Tools

• Custom tools allow you to extend the AI Agent’s capabilities by connecting it to your own external systems and endpoints
• Click Add New Tool to create a new tool and fill in the following sections\
  

​

Basic Information

• Give your tool a name using snake_case with no spaces — the AI sees this name when deciding which tool to call
• Add a clear description so the AI knows exactly when to use this tool
• Select your HTTP method and provide the endpoint URL the AI will call

Tool Name: fetch_customer_details
Description: Use this tool to fetch the details of a customer when they
reach out, so the AI can personalise its response based on their account
and history.
HTTP Method: GET
Endpoint URL: https://api.yourcompany.com/customers/details

​

Authentication

• Configure how the AI will authenticate requests to your endpoint
• Choose from the following options:
  • No Auth — for public endpoints that require no authentication
  • Bearer Token — sends an Authorization: Bearer {token} header
  • API Key — sends a custom header with a key-value pair
  • Basic Auth — encodes username and password as a Base64 header

​

Request Headers

• Add static headers that will be sent with every call to this endpoint
• Authentication headers are managed separately in the Authentication section above

X-Tenant-ID: your-tenant-id
Content-Type: application/json
Accept: application/json

​

Parameters

• Define the inputs the AI will provide when calling this tool
• For each parameter, specify the name, type, location, and whether it is required
• Use available contexts to pass conversation data directly into your tool

Name: order_id
Type: String
Location: Body
Required: Yes
Description: The order ID the customer is asking about

Available contexts:
+ chat_id
+ org_phone
+ chat_name

​

Response Schema (Optional)

• Describe the shape of the API response so the AI can interpret and use the result accurately
• If left disabled, the AI will process the raw response directly

{
  "customer_name": "string",
  "account_status": "string",
  "plan": "string",
  "last_interaction": "string"
}

• Once all fields are filled in, click Create Tool to save and activate it