Dograh is an open-source, self-hostable voice AI platform built under the BSD 2-Clause license as a transparent alternative to proprietary voice agent platforms like Vapi and Retell AI. The core engine is engineered to orchestrate real-time, low-latency audio pipelines, managing bidirectional audio streaming, turn detection, interruptions, and stateful dialog flow. By giving developers full ownership over their voice infrastructure and runtime logic, Dograh eliminates third-party vendor lock-in and per-minute platform markups while safeguarding data residency and user privacy.
At the center of Dograh is an interactive, visual drag-and-drop workflow canvas that allows engineers and product teams to design complex conversational logic without hardcoding state machines. The platform follows a Bring Your Own Key (BYOK) architecture across the entire voice stack: developers can freely mix and match Large Language Models (such as OpenAI, Anthropic, or local open-weights LLMs), Speech-to-Text engines (including Deepgram, Whisper, and AssemblyAI), and Text-to-Speech synthesizers (such as Cartesia, ElevenLabs, and PlayHT) to match their latency, cost, and voice quality requirements.
Dograh provides built-in connectors for enterprise telephony and communication protocols, supporting direct SIP trunking and cloud communications providers like Twilio, Vonage, Telnyx, and Plivo for inbound and outbound calling. Crucially, Dograh is natively integrated with the Model Context Protocol (MCP), enabling voice agents to invoke external tools, query databases, trigger API actions, and collaborate with coding agents (like Claude Code and Cursor) in real time during live voice interactions.
Deployment is straightforward for production and self-hosted environments. Dograh ships with ready-to-run Docker and Docker Compose configurations that can be provisioned on any standard Linux VPS, on-premises server, or private cloud cluster in minutes. The self-hosted edition has zero platform licensing fees, requiring operators to pay only for their underlying compute and third-party provider API keys. A managed cloud option is also available for teams that prefer a hosted setup.
