The voice-AI platform you can actually own

What it does Dograh is a voice-agent platform for building inbound and outbound AI callers. It bundles a drag-and-drop workflow editor, telephony hooks (Twilio, Vonage, Telnyx), and pluggable LLM / STT / TTS backends into a single Docker-deployable stack. The pitch is simple: zero-to-talking-bot in minutes, but with source code you can modify and infrastructure you control.

The interesting bit The project leans hard into “no lock-in” as a feature, not just a slogan. It ships with auto-generated API keys and a default model stack so you can test without configuring anything, then lets you swap in your own providers or run entirely on-prem. The built-in QA node that audits prompt quality across workflow steps is a nice touch—most platforms leave that to external evals.

Key highlights

• BSD 2-Clause license; Python-based, Docker-first deployment
• Visual workflow builder with test mode and in-dashboard web calls (no telephony setup needed to iterate)
• Bring-your-own-key for LLM, STT, TTS, and telephony, or use the bundled stack
• Supports call transfers to human agents; MCP-native architecture
• Python and Node SDKs available; maintained by YC alumni

Caveats

• English support only at present; other languages are marked “expandable” but not yet documented
• Anonymous telemetry is on by default; you must explicitly opt out via ENABLE_TELEMETRY=false
• The “under 2 minutes” claim assumes local Docker; remote HTTPS deployment involves more steps

Verdict Worth a look if you’re building voice agents and need data residency or want to escape per-minute SaaS pricing. Skip it if you need mature multilingual support or want a fully managed, hands-off service.