Skip to main content
MCP tools follow a resource_action naming convention and map 1:1 to the documented REST endpoints. The scopes on your API key decide which tools are callable — read tools need :read, mutations need :write, and deletions need manage.
The authoritative, always-current list is whatever the server returns from tools/list — ask your client to “list available Agntix tools.” The catalog below is the stable set grouped by resource.

Agents

Chat

Voice

Tools

Phone Numbers

Analytics & account

These map to the OpenAPI-generated resources under More Resources.

Example interaction

A single natural-language request can fan out into several scoped tool calls:
You: “Create a concise support agent on gpt-4o-mini, then send it a test message.”
  1. agents_create → returns agnt_…
  2. chat_session_create with that agentId → returns ses_…
  3. chat_message_send → returns the assistant’s reply
Each call is authenticated and authorized exactly as the equivalent REST request would be.