Add voice to your agent

Cloudflare introduces an experimental voice pipeline for the Agents SDK that enables real-time speech input and TTS output within the existing Durable Object-based agent architecture. The post documents APIs and utilities (withVoice, withVoiceInput, useVoiceAgent, VoiceClient), built-in Workers AI STT/TTS providers, client and server code examples, support for telephony adapters (e.g., Twilio), transport options, hooks for customization, and guidance for integrating third-party providers; it is a developer-focused feature announcement rather than a security advisory.