Documentation
How to build voicast into your product: conversational voice AI for phone calls and the Web, through the API, MCP and SDK.
voicast is a conversational AI platform for developers. Your product (typically a SaaS) brings the phone numbers and the customers; voicast answers the call, greets the caller, collects what you ask for, answers frequent questions and hands off to a person when needed. After the call you get the outcome, the collected fields, a transcript and a summary through the API and webhooks.
You keep your phone numbers, carrier and customers. voicast keeps the agents (what the AI says and asks), the tenants (the businesses that take the calls) and the call records.
Pages#
| Page | What it covers |
|---|---|
| Getting started | From sign-up to your first call |
| Agents | Role, greeting, fields to collect, FAQ, handoff rules, voice, speaking rate |
| Phone | Twilio Media Streams (TwiML), cloud PBX and contact-center systems, handing off to a person |
| Web and apps | Sessions and WebRTC, audio or text output, pairing with avacast |
| Results | Calls, transcripts, summaries and fields; webhooks and signature verification |
| Tools | Letting the AI call your API during a call (signatures, confirmation, timeouts, idempotency) |
| Connections | Presets such as Google Calendar, kintone and Slack, and custom connections |
| MCP | Operating voicast from Claude, Claude Code, Cursor, ChatGPT and other agents |
| SDK | The Node.js / TypeScript SDK |
| Pricing and billing | ¥6 per minute, prepaid credit and auto-recharge, enterprise contracts |
| Errors and limits | Status codes, error codes, rate and size limits |
| API reference | Every endpoint |
Endpoints#
| Purpose | URL |
|---|---|
| API | https://api.voicast.jp/v1 |
| Dashboard | https://app.voicast.jp |
| Phone audio (Twilio Media Streams) | wss://voice.voicast.jp/ws |
| Web and app audio (WebRTC) | https://voice.voicast.jp/v1/webrtc/offer |
| MCP | https://api.voicast.jp/mcp |