xAI launches no-code voice agent builder at $0.05/min
By Steven Van ·
Beta on console.x.ai/voice/agents wires Grok Voice, telephony, retrieval and guardrails into one speech-to-speech stack.
xAI has opened the beta of its Voice Agent Builder, a no-code platform for building human-like voice agents on top of Grok Voice. Instead of stitching together separate speech-to-text, language model, and text-to-speech providers, you get one interface built directly around Grok Voice, with pricing that starts at $0.05 per minute of audio.
One speech-to-speech stack instead of three
The pitch is architectural. Most voice agents today are assembled from three vendors: an STT service to transcribe the caller, an LLM to reason, and a TTS service to speak back. Every hop adds latency and a new failure point. xAI says Voice Agent Builder runs on a single speech-to-speech path tightly coupled to Grok Voice, which is how it claims sub-second latency in real VoIP production environments. You configure one interface rather than gluing three APIs together and hoping the handoffs stay in sync.
Everything a production agent needs, in the box
The builder ships with the surrounding stack pre-wired: telephony, knowledge retrieval, tools, guardrails, and observability. You start by writing a plain-language description of how calls should flow, then attach your documents, tools, and guardrails. The retrieval layer accepts common formats such as plain text, Markdown, Word, PowerPoint, Excel, HTML, and JSON, and the agent pulls from those files live during a call. Crucially, it works with your existing phone numbers, APIs, and MCPs, so a voice agent can trigger real actions rather than just talk. xAI says you can go from zero to a working agent in about two minutes.
Pricing built to undercut the incumbents
Voice Agent Builder is priced at $0.05 per minute of agent audio, with an additional $0.01 per minute for telephony on provisioned phone numbers. The phone numbers themselves come at no extra charge, and every account gets a free number to start. For enterprises, direct SIP connectivity lets agents plug into existing telephony infrastructure. That per-minute rate is aimed squarely at cost-conscious buyers and reportedly lands below established voice-agent players. The platform supports 25+ languages and 80+ distinct voices, building on the voice cloning xAI shipped earlier in 2026.
Who this is for
The obvious fit is high-volume phone work: support lines, appointment booking, screening and qualification, and outbound follow-ups where a team wants a production agent without building and maintaining the plumbing. Because it is no-code and beta-available today at console.x.ai/voice/agents, operators can test a live agent on a real number the same afternoon, then layer in guardrails and observability before scaling. If you already use Grok elsewhere in your stack, this keeps the voice layer on the same model.
Sources: xAI — Introducing the Voice Agent Builder, xAI on X, Crypto Briefing.