What is Dograh?
Dograh is an open-source, self-hostable voice agent platform and workflow builder designed as an alternative to proprietary services such as Vapi and Retell. Licensed under the BSD 2-Clause license, it allows teams to construct, customize, and manage conversational voice pipelines. The platform supports standard cascade architectures consisting of Speech-to-Text (STT), Large Language Models (LLM), and Text-to-Speech (TTS), as well as direct, single-hop Speech-to-Speech (S2S) models. Dograh can be deployed entirely on-premises, within a customer's private cloud, or accessed through a managed cloud service.
Dograh key features
- Supports both modular cascade workflows (STT, LLM, TTS, and telephony) and native speech-to-speech model pipelines.
- Enables on-premises, virtual private cloud (VPC), and air-gapped deployments to keep call audio, transcripts, and inference within local network boundaries.
- Includes a Model Context Protocol (MCP) server that allows developers to create, modify, and deploy voice agents directly from coding environments like Claude Code and Cursor.
- Integrates with external APIs as well as self-hosted open-source models, including Whisper, Voxtral, and Kokoro.
- Blends pre-recorded human audio clips with live text-to-speech using identical voice profiles to minimize latency and token usage.
- Provides multiple operational models, spanning self-hosted open-source deployments, fully managed cloud hosting, and managed VPC installations.
Who is Dograh for?
Dograh is intended for software developers, voice application engineers, and enterprises that require full data sovereignty and architectural flexibility over their voice automation stack. It is tailored for organizations in regulated industries—such as healthcare, banking, fintech, insurance, legal intake, and government—that must adhere to data protection standards like HIPAA and GDPR by preventing customer audio and personally identifiable information from exiting their private infrastructure.