When we started building Mangal, the easy path was a closed, all-in-one voice platform: plug in the vendor's models, ship the dashboards, done. We chose the harder path. Mangal is built on Dograh, an open-source voice AI stack, and we contribute back to it. This isn't a technical footnote — it changes what customers can do with the product.
Open source means you're not locked to a black box
With a closed stack, the answer to "how does this actually work?" is a support ticket. With an open stack, it's the code. That matters for teams whose calls involve sensitive data, compliance review, or unusual workflows: they can inspect what happens to their calls, understand the pipeline, and know what they're running in production — rather than trusting a promise.
It keeps the stack honest
An open foundation keeps us honest internally, too. We can't quietly ship a degraded model or a routing change without it being visible. For a product that sits on a business's main phone line, that visibility is a feature, not a cost.
Faster, because we're not reinventing everything
Dograh gives us a battle-tested core — speech-to-text, turn-taking, streaming — so we can spend our energy where it matters for Indian businesses: mixed Hindi–English conversations, WhatsApp as the natural next step, and integrations with the CRMs and calendars our customers already use. Building on open source means standing on the shoulders of the community instead of rebuilding it.
Giving back is part of the deal
Open source only works if the people who use it also contribute. We fix bugs upstream, share improvements that are useful beyond Mangal, and keep the relationship with the Dograh community active. It's the same instinct that runs through the rest of how we work: build in the open, don't hide in a black box.
What it means for you
For customers, this shows up in practical ways:
- Transparency. Every call is recorded with audio and transcript — and the stack underneath is open to inspection.
- No vendor stranglehold. If you ever need to, you're not trapped in a proprietary format or API.
- Real capabilities, not slideware. The voice quality and latency you hear in a demo come from a stack that's tested in the open, not a marketing benchmark.
We don't build in a black box. We build on open source, and we contribute back.
Mangal is the layer on top of Dograh that makes voice AI work for Indian business calling — languages, WhatsApp follow-ups, CRM actions, and human handoff. More about how Mangal started, or hear it on a live call.