Dograh is an open-source voice agent platform built for teams that want to run their own AI calling stack instead of leaning on a closed vendor. Think of it as an open-source take on tools like Vapi or Retell: you can self-host it on your own servers, run it through Dograh’s managed cloud, or have the team deploy it inside your private VPC. You choose your own speech-to-text, LLM, and text-to-speech providers (or run speech-to-speech models like Gemini Flash Live for lower-latency).
To get started with Dograh, you’d typically clone the repo and spin it up with Docker Compose for self-hosting, or sign up for the free-to-try cloud version at app.dograh.com if you’d rather skip the ops work. There’s also an MCP server built in, so you can build and deploy voice agents straight from Claude Code, Cursor, or another MCP-compatible IDE if that’s your preference.
