Skip to content
News Product launch

xAI launches Grok Voice Agent Builder in beta, charges $0.05 per minute with no platform fee

· by Pondero Newsdesk

The short version

xAI released Voice Agent Builder in beta on July 1, 2026, a no-code platform that builds production-ready phone agents from plain-language prompts in under two minutes.

xAI launches Grok Voice Agent Builder in beta, charges $0.05 per minute with no platform fee

xAI released Voice Agent Builder in beta on July 1, a no-code platform that takes a plain-language description of how calls should flow and produces a callable phone agent in roughly two minutes. The pricing structure is flat: $0.05 per minute of audio at the API rate, plus $0.01 per minute for telephony on a provisioned number, with no separate platform fee.

What the platform does

Voice Agent Builder runs on Grok Voice Think Fast 1.0, xAI's speech-to-speech model. Per xAI's announcement, the builder skips the three-component architecture common in competing stacks (separate speech-to-text, language model, and text-to-speech services) in favor of a single audio path tightly coupled to the model.

The builder includes several practical components out of the box:

  • A knowledge base that accepts plain text, Markdown, Word, PowerPoint, Excel, HTML, and JSON documents, organized into reusable collections
  • Tool connectors for Google Calendar, Outlook, Linear, Notion, Google Drive, OneDrive, and arbitrary APIs
  • MCP support for wiring agents to external model-context servers
  • 80+ built-in voices plus voice cloning from roughly two minutes of audio
  • One free provisioned phone number per account, plus Direct SIP trunking for enterprises bringing existing numbers
  • Call recordings and transcripts with tool-use logs

On the benchmark side, xAI's tau-voice Bench leaderboard puts Grok Voice Think Fast 1.0 at 67.3% overall, compared to 43.8% for Gemini 2.5 Flash Live and 35.3% for GPT Realtime 1.5 per xAI's announcement. The benchmark tests agents under conditions xAI describes as real-world: low-quality telephony audio, background noise, accents, mid-sentence interruptions, and multi-language calls.

Why it matters

Phone agent infrastructure has required developers for the past several years. Retell AI, Bland AI, and VAPI each charge separately for voice minutes, the LLM calls behind each turn, and sometimes platform access on top. xAI's single meter collapses that math: a ten-minute support call costs $0.60 total, per the pricing shown on the xAI announcement page.

The no-code interface shifts the practical audience. An operator running a small support line could describe call flows in prose, upload product documentation, and connect a calendar without writing any code. That was not possible with existing voice agent platforms aimed at developers. Whether the prose-to-agent quality holds across complex routing logic is something to judge by calling the agent directly; xAI invites that test on the announcement page.

The MCP connector matters for operators already using agentic workflows. An existing MCP server wired to a CRM or ticketing system can be attached to a voice agent in the same step as the prompt, without re-implementing tool calls in a different SDK.

Context

xAI entered voice AI earlier in 2026 with the Grok Voice Agent API, which let developers build real-time speech interactions over WebSocket. Voice Agent Builder is the no-code layer on top of that API, aimed at operators who do not want to manage audio streaming directly.

The $0.05 per minute price is the current API rate listed on xAI's pricing documentation. Pricing for beta products can change; operators building on the platform should note that the entry price does not lock in future rates.

Competitors Retell AI and VAPI have not publicly responded to the launch. Bland AI has also been quiet. A frontier model company owning the full audio path (training, inference, telephony) is structurally different from platforms that aggregate third-party STT, LLM, and TTS vendors, which gives xAI more room to reprice aggressively if it chooses.

What to watch next

Two things will indicate whether Voice Agent Builder gains traction beyond early adopters: first, whether xAI opens a marketplace for pre-built agent templates that makes onboarding faster for non-technical users; second, how competing voice platforms respond to a foundation-model lab entering at $0.05 per minute with no additional platform tier.

Sources