xAI has introduced Voice Agent Builder, a no-code platform that lets operators and developers configure production voice agents on Grok Voice in under two minutes. The tool, now in beta, targets businesses that want high-volume voice agents without assembling the surrounding infrastructure themselves.
Out of the box, the platform bundles telephony, knowledge retrieval, tools, guardrails, MCPs, and observability into a single interface. Companies can also bring what they already have: existing phone numbers over SIP, custom tools wired to internal APIs and MCP servers, or their own client connected over WebSocket.
Why Voice Agent Builder Is Different
Most voice stacks stitch together three separate APIs: speech-to-text, a language model, and text-to-speech, often hosted by three different providers. Every handoff between them adds cost, delay, and a new point of failure. Voice Agent Builder instead runs on a speech-to-speech path built specifically for Grok Voice, tightly coupled to the model rather than assembled from disconnected parts. This approach reflects a broader shift happening across artificial intelligence tools, where tighter integration between components is replacing loosely connected pipelines.
Trained on the Hardest Calls Available
Real customer calls are messy. They involve poor telephony audio, background noise, heavy accents, interruptions, and callers who change their minds mid-sentence, often across more than 25 languages. xAI trained Grok Voice specifically on calls like these, then built a benchmark called τ-voice Bench to measure performance under identical conditions.
| Model | τ-voice Bench Score |
|---|---|
| Grok Voice Think Fast 1.0 | 67.3% |
| Gemini 3.1 Flash Live | 43.8% |
| GPT Realtime 1.5 | 35.3% |
The benchmark breaks results down further across overall performance, retail, airline, and telecom scenarios, giving businesses a clearer picture of how an agent might perform in their specific industry.
Setting Up an Agent in Two Minutes
Configuration starts with a plain-language description of how calls should flow. From there, users attach documents, tools, and guardrails. An agent begins with a prompt describing call behavior, and the model reasons in real time so it can follow long instructions and handle ambiguous requests without a rigid script.
Knowledge comes from an uploaded knowledge base supporting formats like plain text, Markdown, Word, PowerPoint, Excel, HTML, and JSON. Documents are organized into collections that can be attached to multiple agents at once, keeping policies and product specs centralized instead of duplicated across every prompt.
Tools, Voices, and Telephony
Agents can take action through tools and connectors: scheduling appointments in Google Calendar or Outlook, sending confirmation emails, checking order status through internal APIs, issuing refunds, managing tickets in Linear or Notion, or pulling files from Google Drive or OneDrive. When information isn’t in the knowledge base, the agent can run a web search or an X search for current public information. If a caller needs a human, the agent transfers the call and can notify a team in real time.
Agents can use one of more than 80 built-in voices or a cloned brand voice generated from roughly two minutes of audio. Every account includes a free phone number for testing or production use, and businesses can connect an existing number over SIP from any major telephony provider. Every call is recorded and transcribed, letting teams review audio, read transcripts, and see exactly which tools the agent used during a conversation.
Pricing for the Voice Agent Builder
xAI is billing usage at its standard API rate of $0.05 per minute of audio, with voices included and no separate platform fee. Telephony on a free provisioned number costs an additional $0.01 per minute. That contrasts with typical voice stacks, which often bill separately for recognition, reasoning, synthesis, and platform access, each with its own meter. The push for simpler pricing mirrors trends across telecom and cloud services, where bundled billing is increasingly seen as a selling point.
The launch adds to a growing field of voice automation tools aimed at customer service, sales, and support teams, an area closely tied to broader economy shifts as companies automate more of their call center operations. Developers interested in testing the platform can explore xAI’s official site for more details on Grok Voice and the Voice Agent Builder beta.
For businesses evaluating automation tools more broadly, the underlying apps and software ecosystem continues to expand rapidly, with voice-first products increasingly competing alongside traditional chat-based interfaces.
