Introduction
Voice Agents is Sarvam’s end-to-end platform for building voice agents that actually work in India.
It orchestrates a constellation of models (speech recognition, reasoning, and speech synthesis) into one coherent voice agent, with every model in the stack self-hosted by Sarvam. The stack has been hardened on real calls, at scale: messy telephony lines, code-mixed speech, callers who interrupt, languages that switch mid-sentence, and flows that have to hold up across thousands of concurrent calls.
What the platform does
Everything lives in a single workspace. One place to build, test, manage, deploy, run campaigns, and read the data back, with no UIs to wire and no orchestration code to write.
Design agents, then test them in test agent before going live.
Keep agents, versions, and settings organised as your setup grows.
Go live on telephony: bring your own provider or rent a number from Sarvam.
Kick off batch outbound calling at scale.
Call logs, analytics, and custom dashboards over every conversation.
Fetch insights from your conversations and keep improving your agents.
Opinionated doesn’t mean rigid. You still control the prompt, the voice, the tools, and the flow. You just don’t have to build the plumbing underneath them. From a blank agent to a live phone call, everything you need is already here.