Skip to main content

​
Tools

Tools extend your agent’s capabilities beyond conversation. They allow the agent to perform real actions during a call — like booking appointments, looking up customer records, or triggering webhooks.

​
How Tools Work

  1. You define a tool with a name, description, and expected parameters
  2. The tool is attached to an agent
  3. During a call, the LLM decides when to invoke a tool based on the conversation context
  4. The tool executes (API call, webhook, etc.) and the result is fed back to the agent
  5. The agent uses the result to continue the conversation
This is also known as function calling — a capability of modern LLMs.

​
Creating a Tool

  1. Navigate to Build > Tools
  2. Click New Tool
  3. Configure:
    • Name: a clear function name (e.g. check_availability)
    • Description: explain what this tool does — the LLM reads this to decide when to use it
    • Parameters: define the inputs the tool expects (JSON Schema format)
    • Endpoint: the URL to call when the tool is invoked

​
Tool Configuration

FieldDescription
NameFunction name the LLM will reference
DescriptionWhen and why to use this tool (crucial for accuracy)
ParametersJSON Schema defining expected inputs
Endpoint URLHTTP endpoint called with the parameters
MethodHTTP method (GET, POST, etc.)
HeadersCustom headers (e.g. API keys)

​
Example: Appointment Booking Tool

{
  "name": "book_appointment",
  "description": "Book an appointment for the caller. Use when the caller explicitly requests to schedule a visit.",
  "parameters": {
    "type": "object",
    "properties": {
      "caller_name": {
        "type": "string",
        "description": "Full name of the caller"
      },
      "date": {
        "type": "string",
        "description": "Requested date in YYYY-MM-DD format"
      },
      "time": {
        "type": "string",
        "description": "Requested time in HH:MM format"
      }
    },
    "required": ["caller_name", "date", "time"]
  }
}

​
Best Practices

The LLM uses the tool description to decide when to call it. A vague description leads to incorrect or missed invocations. Be specific about trigger conditions.
Your API endpoint should return clear error messages. The agent will relay these to the caller, so make them human-readable.
One tool = one action. Don’t create a “do everything” tool. Split complex workflows into multiple tools with clear responsibilities.
Call your agent and try to trigger each tool. Verify the parameters are passed correctly and the response is used appropriately.