X AI has launched Grok Voice Agent Builder, a no-code platform that enables users to configure production voice agents in under two minutes. The tool is designed for operators and developers seeking to deploy high-volume voice agents without building the surrounding infrastructure from scratch. With Grok Voice, users can access telephony, knowledge retrieval, tools, guardrails, MCPs, and observability in a single interface. The platform also supports integration with existing systems, allowing users to bring their phone numbers over SIP, wire tools to APIs, or connect their own client via WebSocket. Source: xai
Grok Voice is trained on real calls that include low-quality audio, background noise, strong accents, interruptions, and callers who change their minds mid-sentence. These calls often involve ambiguous workflows that span dozens of tools and occur in 25+ languages. τ-voice Bench measures agents under the same conditions, with Grok Voice achieving a 67.3% score in the Think Fast benchmark. Other models, such as Gemini 3.1 Flash Live and GPT Realtime 1.5, scored 43.8% and 35.3%, respectively. Source: xai
The platform allows users to describe call workflows in plain language and attach documents, tools, and guardrails to create a working agent in about two minutes. Agents use a prompt to understand how calls should flow and can follow long instructions or handle ambiguous requests in real time. Users can upload documents in various formats, and the agent retrieves information from these documents during calls. Documents are organized into collections that can be shared across agents to maintain consistency in policies, product specs, and runbooks. Source: xai
