How to Build a Voice AI Agent with OpenAI's GPT-Live, Twilio Agent Connect, and TypeScript
Blog post from Twilio
Twilio’s tutorial explains how to build a TypeScript voice AI agent that connects phone calls to OpenAI’s full-duplex GPT-Live models through the open-source Twilio Agent Connect SDK, avoiding custom WebSocket implementation. After configuring Twilio and OpenAI credentials, a VoiceChannel and Fastify-based server, developers set a GPT-Live session with audio formatting, system instructions, and an optional proactive greeting for inbound callers. The example adds a weather tool through OpenAI Responses API delegation, allowing GPT-Live to decide when to request external information while Twilio Agent Connect executes the TypeScript function and returns its result without blocking the conversation. The tutorial also covers configuring a Twilio phone number webhook, testing calls, logging call transcripts, and initiating compliant outbound calls for uses such as reminders or order updates. It notes that the same architecture can support other messaging channels, alternative voice backends, per-call configuration, recordings, answering-machine detection, and more production-oriented call controls.
Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.