CRM
Read or write customer context through controlled integration boundaries.
Use APIs, webhooks and custom tool boundaries to connect voice workflows with business data and actions.
Organize connections around capabilities such as CRM, calendars, databases, automation, messaging, webhooks, APIs and custom tools.
Read or write customer context through controlled integration boundaries.
Connect scheduling and availability logic to conversational workflows.
Query application data through scoped tools rather than exposing direct database access to the model.
Trigger downstream automation from structured voice events and workflow actions.
Hand structured outcomes to messaging systems where configured.
Receive or emit event payloads at explicit integration points.
Represent MCP-compatible tool surfaces as one possible integration category where supported.
Connect application logic through HTTP-based service boundaries.
Wrap business-specific actions with explicit inputs, outputs and runtime constraints.
A tool should expose a narrow operation with clear inputs and outputs. That keeps the voice agent connected to your application without turning the model into an unrestricted systems layer.
Describe exactly what information an action requires.
Keep business logic inside the application service that owns it.
Return only the context the conversation needs to continue.
// Illustrative tool definition
const updateDeliveryWindow = {
name: "update_delivery_window",
input: { orderId: "string", window: "string" },
execute: async (input) => app.updateWindow(input)
};
Start with clear integration boundaries and connect the systems your voice workflows depend on.
Developer Overview