Skip to main content
Custom mid-call tools let your AI assistant interact with your external systems during a call — checking inventory, verifying customer data, fetching real-time information, or triggering any API you own.
No coding is required to configure a custom tool. You define the API endpoint, parameters, and when the AI should use it. The AI automatically knows how to call it.

​
What you can build

  • Look up order status from your backend
  • Verify a customer’s identity against your CRM
  • Check appointment availability from a custom calendar system
  • Fetch product pricing, stock levels, or account information
  • Trigger any REST API endpoint with dynamic data from the conversation

​
Setting up a custom tool

​
1. Create the tool

1

Navigate to mid-call tools

In your assistant settings, go to Prompt & Tools → Custom Mid-Call Tools and click Create Mid-Call Tool.
2

Fill in the main settings

Configure the essential details:
  • Name: Use lowercase letters and underscores (e.g., check_order_status). This is what the AI refers to internally.
  • Description: Explain when and how the AI should use this tool. Be specific — this directly affects when the AI decides to call it.
  • Endpoint: Your API URL (e.g., https://api.yourcompany.com/orders).
  • Timeout: How long to wait for a response, in seconds.
  • Method: Choose GET, POST, PUT, PATCH, or DELETE.
3

Add headers

Include any required headers, such as:

​
2. Define parameters

Parameters are the pieces of information the AI collects during the call and passes to your API.
1

Add a parameter

For each piece of data your API needs, configure:
  • Name: The parameter identifier (e.g., order_number)
  • Type: string, number, or true_false
  • Description: What the AI should collect and any format requirements
2

Set validation rules

Add format requirements in the description field:

​
Parameter types

String

Text values like names, addresses, or reference numbers.

Number

Numeric values like amounts, quantities, or IDs.

true_false

Boolean values for yes/no situations.

Format hints

Add format instructions in the description to guide the AI.

​
Dynamic endpoints

You can make your endpoint URL dynamic by including parameter names in curly braces:
The AI automatically replaces {order_id} with the actual value it collects during the conversation.
Enclose variable names in curly braces and use the exact parameter name as defined in your parameter configuration.

​
Testing your tool

Click Test Tool to run a test with dummy data:
  • String parameters use "test"
  • Number parameters use 1
  • Boolean parameters use true
You’ll see the response status code and body, so you can verify your endpoint is reachable and returning the expected structure before going live.

​
Integrating with the automation platform

For complex logic that goes beyond a single API call, connect your custom tool to the no-code automation platform:
1

Create a webhook flow

In the automation platform, create a new flow with a Webhook trigger, add your logic (multiple API calls, data transformation, error handling), and end with Return Response.
2

Use the webhook URL as your endpoint

In your custom tool settings, set the endpoint to your webhook URL and append /sync:
This approach lets you transform data, make sequential API calls, apply business logic, and handle errors gracefully — all without code.

​
Real-world examples

The AI will:
  1. Ask the caller for their order number.
  2. Fetch the status from your API.
  3. Explain the delivery date and status in plain language.
The AI will:
  1. Ask about the desired service.
  2. Get the caller’s preferred date.
  3. Show available time slots.
The AI will:
  1. Collect the caller’s contact details.
  2. Verify them against your CRM.
  3. Proceed based on the verification result.

​
Configuring your assistant to use the tool

Add instructions to your system prompt so the AI knows when and how to use each custom tool:
Test your tools with various conversation flows before going live. Start with simple scenarios and work toward edge cases.