Skip to main content

Tools Overview

Tools extend your voice agent beyond conversation. Without tools, your agent can only speak from its training and the context you give it in the system prompt. With tools, it can query live data, book appointments, update records, and take real-world actions while the caller is on the line.

What tools are

A tool is a function the AI agent can invoke during a live conversation. When the agent decides it needs external data or needs to perform an action, it calls the tool, waits for the response, and then continues the conversation — all in real time, without interrupting the natural flow of the call. Think of tools as the agent’s hands: conversation is how it thinks and speaks, tools are how it acts.

How tools work

1

Agent recognises the need

Based on the conversation context and the tool’s description, the AI determines that a tool call is appropriate — for example, the caller asks to book an appointment.
2

Tool is invoked

Truedy sends a request to the tool’s configured endpoint (your server for custom tools, or the built-in integration for native tools), passing any parameters the agent extracted from the conversation.
3

Response is received

Your endpoint returns a JSON response. Truedy injects the result directly into the conversation context.
4

Agent continues

The agent reads the result and responds naturally — “I’ve booked you in for Tuesday at 2 PM” — without the caller ever knowing a system call happened.
Tool calls are non-blocking from the caller’s perspective. The agent will typically acknowledge the action (“Let me check that for you…”) while the tool call completes in the background.

Two categories of tools

Built-in Tools

Pre-built integrations with popular services. Configure your credentials, enable the tool, and your agent is ready. No code required.Available integrations:
  • Cal.com calendar booking
  • GoHighLevel CRM
  • Calendly scheduling
  • End Call (always available)

Custom HTTP Tools

Define a webhook endpoint on your own server. Truedy calls it when the agent needs data or needs to take an action. Connect to any API, database, or internal system.

When to use tools

Tools are the right choice when your agent needs to go beyond static knowledge:
Use caseTool type
Check calendar availability and bookCal.com or Calendly built-in
Look up a contact in your CRMGoHighLevel built-in or custom
Verify a customer’s account statusCustom HTTP tool
Check order history or shipping statusCustom HTTP tool
Create a support ticketCustom HTTP tool
Send a follow-up emailCustom HTTP tool
Update a lead stage after a callGoHighLevel built-in or custom

Tool configuration is per-agent

Each agent has its own list of enabled tools. The same tool (for example, a customer lookup webhook) can be attached to multiple agents, or you can create agent-specific variants with different configurations. Tools are managed in the Tools section of the dashboard, and can also be configured via the API.

The tool description is critical

The AI decides when to call a tool based on its description field. A vague description leads to missed calls or inappropriate calls. A precise description leads to reliable behaviour.
Write tool descriptions from the agent’s perspective: “Call this tool when the caller asks about their order status and you have their phone number.” The more specific you are, the more reliably the agent will use the tool at the right moment.

Response time matters

Tools are called synchronously during a live conversation. Slow tool responses create awkward pauses.
  • Target: respond in under 3 seconds
  • Maximum tolerated: 5 seconds before the agent may continue without the result
  • If your backend operation is slow, return a queued status immediately and instruct the agent to tell the caller to expect a follow-up
If your tool endpoint does not respond within the timeout window, Truedy will not retry — the agent will continue the conversation gracefully without the tool result. Design your prompts to handle this case.

Tool results are injected in real time

Tool responses are fed directly into the model’s context during the active call. The agent has access to the full response payload and can reference any field within it in its reply. There is no post-processing step — what your endpoint returns is what the agent reads.

Next steps

Built-in Tools

Set up Cal.com, GoHighLevel, or Calendly with no code

Custom Tools

Build a custom webhook tool and connect it to any system