NexaVoxa

NexaVoxa is aimed at teams that treat phone calls as a production channel rather than a demo. The build flow has five stages: name the agent and choose whether it runs from a single prompt or a branching guided flow; attach the pages, documents, and recordings it should answer from; define the tools it can call mid-conversation and the data it should capture; test the agent by voice in the browser and read the live transcript; then connect a Twilio, Telnyx, or SIP trunk and route numbers to it. The same agent can also be embedded on a website or dialed out from a contact list.

Because the infrastructure layer is the product, provider choice stays open. Agents can run on models from OpenAI, Gemini, or Azure, with separate selections for language model, speech-to-text, and voice. Pricing is transparent about this: the pricing page exposes a per-minute cost calculator, and the published example stacks GPT-4o Mini, Deepgram Nova-3, and Deepgram Aura-2 at roughly $0.079 per minute, plus a flat per-call platform fee regardless of call length. Outbound campaigns validate every number before dialing and resume automatically after a restart, which matters for high-volume calling where a mid-run failure would otherwise waste the list.

Operationally, the workspace is built around auditability. Calls are recorded, transcribed by speaker, scored for sentiment, and filed with duration, outcome, and cost. Dashboards expose volume, latency, success rate, sentiment, and spend across every agent and number. The company also maintains a broader product family alongside the voice platform, including call QA and intelligence, HRIS, recruiting, meeting translation, and exam preparation tools, and an enterprise tier adds a dedicated architect, private node deployment, a 99.95 percent uptime SLA, SSO/SAML, and priority support.