According to xAI, users can define how calls should be handled, connect company knowledge bases and external tools, deploy agents using a phone number or SIP connection, monitor conversations, and choose from synthetic or cloned voices.
The company noted that most voice AI applications combine separate speech recognition, language model, and speech synthesis components, with each step adding cost, latency, and potential points of failure. Voice Agent Builder instead uses the company’s Grok Voice speech-to-speech model through a single interface, which xAI says reduces latency and simplifies deployment.
The platform supports multilingual conversations across more than 25 languages and language switching. Pricing starts at USD 0.05 per minute, with additional charges for services including telephony using xAI phone numbers.
The launch follows a series of voice AI announcements from xAI over the past months, beginning with the Grok Voice Agent API introduced in December 2025, when xAI opened its in-house voice technology to developers to build voice agents. Since then, the company has expanded its offering with speech-to-text and text-to-speech APIs, conversational voice models, custom voices, and now a visual environment for designing and deploying voice agents.
With this launch, xAI joins a group of companies offering platforms for building and deploying enterprise AI voice agents, including ElevenLabs, Synthflow, Retell AI, PolyAI, and Vapi.