# Vaani Developer Platform: Voice Agents, Pricing and What to Check

Vaani's self-service launch brings call-flow controls and a published per-minute price to teams building voice agents.

By Paras Madan | 2026-10-30


Vaani has launched a self-service developer platform for voice AI agents, giving teams a way to configure conversations rather than commission a bespoke deployment. The useful question is how much control developers get when a call stops following the script, and what they pay while finding out.

## What Vaani launched

Vaani founder Nitesh Tripathi set October 29 as the Developer Platform launch date in a [post shared by the company](https://in.linkedin.com/company/vaaniresearch). His explanation focused on the difficult parts of voice agents: call flow, interruptions and edge cases. A polished voice alone does not solve those problems.

The [official product site](https://vaaniresearch.com/) now links to the agents platform and describes a self-service studio with a visual graph builder, prompt management, call review and evaluations. Developers can also work programmatically with agents, voices, conversational graphs and call data.

This changes the starting point for a small product team. Instead of treating a voice agent as an integration project first, it can begin with a narrow conversation and test the branches that make it useful. That does not remove the integration work. It makes the conversation itself easier to inspect before committing to a wider deployment.

## Vaani voice-agent pricing

Vaani advertises its developer stack at **₹5 per minute, described as all-in**. It separately lists standalone text-to-speech at ₹2.5 per 1,000 characters. These are different products and billing units, not two interchangeable prices for the same call.

At the listed minute rate, 1,000 minutes would cost ₹5,000. That is arithmetic from the advertised rate, not a tested invoice or a promise about a particular account's final bill. Buyers should confirm taxes, telephone-number charges, destination coverage and any other applicable conditions before budgeting a production rollout.

The more useful comparison is cost per completed task. A cheaper minute can still produce an expensive workflow if callers repeat information, abandon the call or need a person to restart it. Track completion, escalation and repeated attempts alongside the duration.

## What to test before connecting real customers

Start with a workflow whose successful outcome is easy to verify, such as collecting a preferred appointment time. Build a test set containing corrections, interruptions, silence and an answer that does not fit the expected options. Then check whether the system preserves the caller's latest intent through those changes.

The [developer documentation](https://docs.vaanivoice.ai/) is the entry point for implementation. For evaluation, save both the intended outcome and the actual result of each test. A fluent transcript can conceal a wrong action, so listening to the call is only one part of checking it.

One particularly useful test is a caller changing a number halfway through a sentence. If an agent captures the first number confidently and ignores the correction, better voice quality will not fix the business error.

## A platform launch is not a complete model release

Vaani's site lists voice synthesis as active, while voice recognition and voice reasoning remain in development. The self-service launch should therefore not be read as the release of every component in its research roadmap. Likewise, vendor latency claims are not a substitute for testing the languages, network conditions and integrations a deployment will use.

For teams building voice products, the opportunity here is a shorter path from a conversation idea to an inspectable workflow. The next thing to watch is whether that control survives real callers, where the important improvement is often a correctly handled interruption rather than a more impressive demo.

I'm Paras. I work at the intersection of AI, engineering and distribution.

I help technical teams turn launches and complex product workflows into clear stories developers can understand and share.

[Book a 30-minute call](https://cal.com/paras-madan/30min).