Dograh: The Open-Source Voice AI Platform That Could Kill Your VAPI Bill

By Prahlad Menon 4 min read

If you’re building voice AI applications, you’ve probably felt the sting of per-minute SaaS pricing from platforms like Vapi or Retell. A few hundred test calls during development, a demo that runs long, and suddenly you’re looking at a bill that makes you question your life choices.

Enter Dograh — an open-source, self-hostable voice agent platform that promises unlimited voice agents for exactly $0.

What Is Dograh?

Dograh is a drag-and-drop voice agent builder that you can run on your own infrastructure. It’s BSD 2-Clause licensed (the same license as FreeBSD), which means you can use it commercially, modify it, and distribute it without worrying about copyleft requirements.

The pitch is simple: from zero to a working voice bot in under 2 minutes.

curl -o docker-compose.yaml https://raw.githubusercontent.com/dograh-hq/dograh/main/docker-compose.yaml && \
  REGISTRY=ghcr.io/dograh-hq docker compose up --pull always

That’s it. Open http://localhost:3010 and you’re building voice agents.

Why This Matters

Here’s the honest comparison that caught my attention:

FeatureDograhVapiRetell
LicenseBSD 2-ClauseProprietaryProprietary
Self-hostableYesNoNo
PricingFree (self-host)Per-minutePer-minute
Bring your own LLM/STT/TTSAny providerTheir integrationsTheir integrations
Source customizationFull accessNoneNone
Data residencyYour infraTheir cloudTheir cloud

The kicker: no vendor lock-in. If Dograh disappears tomorrow, you still have the code. If Vapi triples their prices, you’re stuck renegotiating or rewriting.

Key Features

Zero-config start: Ships with auto-generated API keys and its own LLM/TTS/STT stack. Connect your own keys later if you want.

Telephony built-in: Twilio, Vonage, Vobiz, Cloudonix integrations out of the box. Transfer calls to human agents when the bot gets stuck.

Test mode: Try your agent end-to-end before publishing, with no production calls affected. Make web calls directly from the dashboard while building.

QA Node: A workflow node that analyzes prompt quality across your other nodes. It’s like having a second pair of eyes on your prompts.

AI Personas for Testing: This is clever — Dograh ships with AI personas that mimic real customers to test your bots. No more asking your coworker to pretend to be an angry customer for the fifth time today.

Who’s Behind It?

Built by YC alumni and exit founders. They claim to be “committed to keeping voice AI open” — time will tell, but the BSD license is a strong signal. They can’t rug-pull the open-source version.

When Would You Use This?

Good fit:

  • You’re building voice agents and want to control costs
  • You need data residency (healthcare, finance, government)
  • You want to customize the underlying code
  • You’re prototyping and don’t want to burn through per-minute credits

Maybe not:

  • You need enterprise SLAs and support contracts
  • You’d rather pay than maintain infrastructure
  • You’re building a quick demo and don’t care about long-term costs

The Catch?

Self-hosting isn’t free in the “zero effort” sense. You need to run Docker somewhere, handle updates, and debug issues yourself. The community Slack seems active, but you won’t get a support ticket escalation path.

Also, first startup takes 2-3 minutes to download all images. Not a dealbreaker, but worth knowing.

Bottom Line

Dograh looks like a serious contender in the voice AI space. The BSD license, self-hosting option, and bring-your-own-everything approach make it worth evaluating if you’re currently paying per-minute for voice agent infrastructure.

For clinical trial screening, patient intake, appointment scheduling — anywhere you’re running voice agents at scale — the math gets compelling fast.

Links: