positioning · OSS economics
Stop paying platform fees._
The core voice stack is a commodity. STT, TTS, and LLM APIs are cheap and getting cheaper. So why does the closed platform on top take 60 to 80 percent of your bill? Bring your own keys, run open source, keep the difference.
§ the UI tax
You're paying 3 to 5x the actual AI cost for a dashboard.
On a typical closed voice AI platform, STT, TTS, and LLM together account for roughly 10 to 20 percent of your bill. Telephony adds another 5 to 10 percent. The remaining 60 to 80 percent is platform margin, charged for the workflow builder and the hosted UI.
That math made sense when the AI primitives were exotic. They're not anymore. When you bring your own keys to an open-source stack, you pay wholesale for the AI, your telephony provider for the calls, and nothing for the dashboard.
§ what you keep
Real ownership, not a marketing checkbox.
§ how it works
Four levers, all in your hands.
Bring your own STT and TTS keys
Connect Deepgram, AssemblyAI, ElevenLabs, PlayHT, OpenAI, or anything else with an API. STT and TTS routing is production-ready. You pay the provider directly at wholesale rates.
LLM swapping, in beta
Plug in OpenAI, Anthropic, Groq, or local models. The LLM swap layer is currently in beta, so it's great for experimentation today and getting steadily more boring for production.
Self-host means no per-minute rent
Run the entire stack on your own infrastructure. The codebase is BSD 2-Clause, so there's no usage cap and no per-minute platform fee. Your bill is your AI providers, your telephony, and your servers.
Full code audit
Every line is on GitHub. Security teams can audit it. Engineering teams can fork it. There is no closed core, no proprietary binary, no surprise terms in the fine print.
Deploy anywhere
Managed cloud if you want to ship today. Self-hosted on your VPC if compliance requires it. Same codebase, same workflow builder, your choice of where the bytes live.
Zero exit cost
Export prompts, call logs, transcripts, and configs whenever you want. Migrate providers without breaking workflows. The whole stack is designed to be leave-able.
§ questions worth asking
The things you're wondering.
On closed voice AI platforms, the 60 to 80 percent of your bill that goes to the vendor on top of STT, TTS, and LLM costs. The underlying AI is the same commodity API everyone else uses. The premium pays for the dashboard. With Dograh and your own keys, you pay only the wholesale AI cost.
§ stop renting your stack
Own the agent. Own the data. Own the bill.
Fork the codebase, bring your keys, and run voice agents without paying a dashboard tax. Or talk to us about migrating off your current platform.