Trigger actions in your own systems during a call

The External Action skill allows your AI agent to make real-time API calls during an active conversation.

This is useful when your agent needs to fetch or send information instantly, for example checking an order status, verifying a booking, or calculating a price while still on the call.

By connecting your external systems through an API, the agent can access up-to-date data and respond immediately, keeping the conversation dynamic and accurate.

External Action is the same skill as the During Call mode of the Call API skill. For the full reference, including Results and Path, see Look up live data during a call.

When to use it

Use this skill when your agent needs to:

• Retrieve or update data in real time while speaking to the caller.
• Perform an action based on what the customer says (for example check shipment status or verify a membership).
• Provide live information from your system, such as pricing, product availability or account details.

How it works

The skill performs an API request during the conversation, triggered by logic you define in the agent's Main Objective.

You can configure:

• API method: choose between GET or POST.
• Parameters: pass information the API needs (like an order ID or email).
• Headers: include authentication or custom headers (for example Authorization: Bearer token).
• Results: specify which values the AI should read and use from the API's response.

The agent then calls your API, processes the result, and uses it naturally in its response to the caller.

Setup steps

1. Go to Agent → Skills → External Action.
2. Fill out the following fields:
   • Action Name: give the action a clear name (for example check_order_status).
   • Description: describe what this action does (for example "Fetch shipment tracking information").
   • API Endpoint: enter the URL of the API.
   • Method: choose GET or POST.
   • Parameters (optional): define input values your API needs (for example order_number, email).
   • Results (optional): define the expected response keys (for example status, estimated_delivery_date).
   • Header (optional): add any required headers, such as authorization tokens.
3. Toggle Enable Skill to activate it.
4. Click Create to save.

Example prompt instruction

Because the External Action runs during the call, you must include a clear instruction in your Main Objective to tell the AI when to trigger it. For example:

If the caller wants to check the delivery status of their order, use the external action check_order_status with the parameter order_number that the caller provides. Then tell them the current status and estimated delivery date.

This ensures the agent knows exactly when to call the API and how to use the result.

Example use case: shipment tracking

Scenario: a customer calls to ask, "Where is my package?"

Action configuration:

• Action name: check_order_status
• API endpoint: https://api.shippingcompany.com/tracking
• Method: GET
• Parameter: order_number (Text)
• Result: status, estimated_delivery_date

Prompt instruction example:

If the caller asks for their delivery status, call the external action check_order_status. Use the result fields {{status}} and {{estimated_delivery_date}} to answer the customer clearly.

Conversation example:

Caller: "Can you tell me where my parcel is?"

AI agent: "Sure, could you please provide your order number?"

(Agent calls the API in real time.)

AI agent: "Your package is currently in transit and should arrive on October 24th."

Other examples

• Membership verification: "If the caller mentions their membership ID, use check_membership_status to confirm whether their account is active."
• Price lookup: "When the caller asks for a quote, trigger the action get_price_estimate using the product name and quantity."

The External Action skill gives your AI agent live access to your systems, allowing it to fetch information, verify details or perform real-time actions while maintaining a smooth, natural conversation.