Dograh is an open-source voice AI platform (BSD 2-Clause) you can self-host with Docker or run on managed cloud. Primary sources describe inbound and outbound agents, cascade STT→LLM→TTS or speech-to-speech, a visual workflow builder, browser test audio, telephony integrations, and MCP for IDE agents. Fields without a primary-source answer are omitted.
Best fit
Teams that want a Vapi/Retell-class voice agent platform with self-hosting, on-prem data control, and BYOK models.
Watch-outs
- Repo topics note Pipecat under the hood; treat Dograh as the product/platform layer, not a drop-in Pipecat replacement.
- Managed cloud concurrency and full STT/TTS/LLM matrices change often—confirm on docs and pricing.
- speech.dev has not independently verified latency, WER, or compliance claims.
