SvaraCall - Subprocessor List
Last updated: 4 October 2026
Overview
Sarwagyna Private Limited Last Updated: 4 October 2026 Version 1.1
This page describes the categories of third parties we engage to help deliver the SvaraCall platform. It is incorporated into our Privacy Policy and Data Processing Addendum.
We give thirty (30) days' notice before adding or replacing any Subprocessor that processes call content or contact data. To receive notifications, email privacy@sarwagyna.com with the subject line “Subprocessor notices”.
What they process
| Category | What they process |
|---|---|
| Authentication | Account login sessions (Supabase) |
| Payments | Credit top-ups and hire-fee checkout (Razorpay) |
| Telephony | Call routing, number provisioning, and KYC for purchased numbers |
| Speech-to-text / text-to-speech | Converting caller speech to text and AI responses to voice, with automatic cross-provider failover if one is degraded |
| Language models | Generating your AI employee's conversational responses from your script and training material, with automatic cross-provider failover |
| Product analytics | Understanding where new users get stuck while setting up their first AI employee (PostHog, EU region). Limited to the setup flow only — it is never active on the screens that show your leads, calls, recordings or transcripts. |
| Storage | Encrypted storage of call recordings |
A named list of current sub-processors is available on request — email support@svaracall.com.
Affiliates
None at present.
Data residency
| Data category | Where it is processed | Notes |
|---|---|---|
| Account login sessions | Supabase | Authentication sessions only |
| Payments | Razorpay | Card data is held by the payment gateway, not by us |
| Product analytics | PostHog, EU region | First-employee setup flow only. Not active on screens that show leads, calls, recordings, or transcripts |
| Call recordings | Encrypted storage | Stored encrypted when recording is enabled |
| Speech-to-text, text-to-speech, and language models | May be outside India | Automatic failover can route a degraded provider to an alternate provider |
Current honest position: SvaraCall is not a fully in-country stack. Speech recognition, speech synthesis, and language-model inference may process data outside India. Customers requiring full in-country processing should contact us before contracting — we cannot offer it on the standard platform today. Named providers and their processing locations are in the list available on request.
Model training
We do not use identifiable customer call audio, transcripts, or contact data to train general-purpose AI models. Where a Subprocessor’s default terms permit training on submitted data, we configure the no-training or zero-retention option where the provider offers one. Per-provider controls are included in the named list available on request.
Change log
| Date | Change |
|---|---|
| 4 October 2026 | Published processing categories. The named sub-processor list is available on request from support@svaracall.com. |
| 14 September 2026 | Initial publication |
Sarwagyna Private Limited | privacy@sarwagyna.com | +91 6305036991