Custom HTTP

Build custom HTTP tools to connect your voice agent to any external API or service.


Overview

Custom HTTP tools allow you to define API calls that your agent can execute during voice conversations. This is useful for integrating with internal systems or APIs not available in the marketplace.

Create custom voice tool modal

Creating a Custom Tool

Navigate to your agent’s Tools tab, click Add Tool, and select Custom HTTP.


Configuration Fields

FieldDescription
Tool NameEnter a name for your custom tool
When to ExecuteDescribe when this tool should be executed. Be specific about the conditions that should trigger this tool
API ConfigurationEnter the API endpoint URL and select the HTTP method (GET or POST)
Message to SpeakMessage to be spoken when executing the tool
HeadersAdd request headers (e.g., Authorization: Bearer {api_token})
Body ParametersDefine the request body parameters

Body Parameters

Each body parameter has:

PropertyDescription
NameThe parameter key name
ModeChoose between Dynamic (LLM determines the value) or Variable (uses a predefined variable)
DescriptionExplain what this parameter does (for Dynamic mode)
RequiredWhether the parameter is mandatory

Dynamic Mode

When set to Dynamic, the LLM determines the value based on the conversation context. Add a description to help the AI understand what value to extract.

Variable Mode

When set to Variable, select from your predefined variables (e.g., {{first_name}}, {{last_name}}). The value will be pulled from the lead’s data.


Example: CRM Update Tool

Tool Name: Update CRM Contact
When to Execute: When the caller provides updated contact information
API Endpoint: https://api.yourcrm.com/contacts/update
Method: POST
Headers:
- Authorization: Bearer {api_token}
Body Parameters:
- first_name (Variable): {{first_name}}
- last_name (Variable): {{last_name}}
- phone (Dynamic): The caller's phone number
- notes (Dynamic): Summary of the call

Best Practices

Be specific in the “When to Execute” field. Clear conditions help the AI know exactly when to trigger the tool.

  • Use descriptive tool names — The AI uses the name to understand the tool’s purpose
  • Provide clear execution conditions — Ambiguous conditions lead to incorrect tool usage
  • Test with sample calls — Verify the tool works before deploying to production
  • Handle errors gracefully — Configure fallback messages if the API call fails

Custom HTTP tools make external API calls during live calls. Ensure your API can respond quickly (under 2 seconds) to maintain natural conversation flow.