Dograh

Dograh is a BSD-2-Clause licensed voice agent platform, positioned as the self-hostable alternative to Vapi and Retell. Agents that answer calls, book appointments, qualify leads or chase payments are assembled in a drag-and-drop workflow builder, tested in the browser, and deployed on your own servers or Dograh's managed cloud.

The defining choice is that every box in the pipeline is swappable. Inbound channel, speech-to-text, language model, text-to-speech and telephony are all bring-your-own-key, so a stack can be re-costed or upgraded without rewriting the agent. There is also a speech-to-speech mode using realtime models that skips the transcription cascade for lower latency, and a hybrid synthesis mode that splices pre-recorded human clips with TTS in the same cloned voice.

Dograh is MCP-native, which means Claude Code, Cursor and other coding agents can create and edit voice workflows programmatically rather than through the canvas — a genuinely unusual authoring path for this category. Python and Node SDKs cover the same ground.

Providers it speaks to. LLMs from OpenAI, Google Gemini, Groq, OpenRouter, Azure, AWS, Hugging Face, xAI, Minimax and Sarvam; STT from Deepgram, AssemblyAI, Speechmatics, Gladia and Dograh's own; TTS from ElevenLabs, Cartesia, Inworld, Rime, Cam.ai and Smallest AI; telephony via Twilio, Telnyx, Plivo, Vonage, Cloudonix and Asterisk; tracing through Langfuse. Whisper, Voxtral and Kokoro cover the fully local path.

Deployment and pricing. Self-hosting is free forever and installs with one Docker command. The managed platform charges $0.01 per minute of platform usage plus roughly $0.07 per minute for models, with no separate platform fee; volume commitments earn model discounts. A fully managed on-prem contract starts at $2,500 per month annually, and a done-for-you build runs $2,000-$6,000. Air-gapped and VPC deployments make it workable for fintech, healthtech, banking and defence teams under HIPAA, GDPR, SOC 2, ISO 27001 or CCPA. The repository has passed 5,000 GitHub stars.