Every channel your customers use, one agent to build

Four: AI phone calls (inbound answered 24/7 and outbound campaigns), WhatsApp through Meta's Cloud API, an embeddable web chat widget, and SIP/BYOC calling on your own carrier. The same agent definition runs on every channel — you build one agent and deploy it across all of them.

No. Zoice connects directly to Meta's WhatsApp Cloud API. You bring your own Meta WhatsApp Business number, connect it through a guided wizard, and pay Meta's conversation rates directly — there is no BSP middleman markup added per message. Zoice bills only for the AI conversation layer.

Yes. Zoice supports BYOC — register your own SIP trunks over UDP, TCP, or TLS with digest or IP-allowlist authentication and SRTP media encryption, then bind your DIDs to AI agents for inbound routing. Every trunk has a test button that sends a real SIP OPTIONS probe to your carrier and reports the result. Plivo number import is live today and Twilio is available as a SIP/provider option.

Attach a knowledge base to the agent and answers are retrieved from your documents. Before going live, the retrieval test bench lets you run a query and inspect the exact top-K chunks the agent would receive, with sources and ranking. Agents append a "Sources:" citation footer to grounded answers, and an LLM-as-judge checks each answer against the retrieved chunks and flags ungrounded ones to a triage dashboard. One knowledge base serves your voice, WhatsApp, and web chat agents.

Every call gets an AI summary, a sentiment verdict with score, structured extraction fields you define, and objective achievement results — plus a downloadable PDF call report and transcript. Conversations are auto-tagged, each one shows its cost in ₹ with a per-minute breakdown, and "Ask Anything" answers plain-English questions over your call data with no query builder or SQL.

10+ Indian languages — Hindi, Tamil, Telugu, Kannada, Malayalam, Bengali, Marathi, Gujarati, Punjabi, and Odia — plus English. Hinglish code-mixing is natively supported, retrieval is language-tagged so a Hindi question returns Hindi chunks, and agents detect a language switch mid-conversation and reply in the language the customer used.