Dograh

Dograh is an open-source platform for building and deploying voice agents, designed to provide organizations with full control over their voice agent infrastructure. It allows users to construct customizable workflows for voice agents by selecting and integrating their own inbound channels, speech-to-text (STT), large language models (LLM), text-to-speech (TTS), and telephony providers. The platform supports both cascade pipelines (STT → LLM → TTS) and real-time speech-to-speech (S2S) audio processing, enabling fluid, natural conversations with features like real turn-taking and interruption handling across more than 70 languages.

A key focus of Dograh is data sovereignty and compliance. The platform can be deployed on-premises or within a private cloud, ensuring that all calls, recordings, transcripts, prompts, and even AI model inference remain inside the organization’s data perimeter. This architecture is designed to meet stringent requirements for data residency and security, supporting compliance with standards such as HIPAA, GDPR, SOC 2, ISO 27001, and CCPA. The open-source nature of Dograh enables organizations to audit the codebase and maintain control over their compliance processes, making it suitable for regulated sectors such as fintech, healthtech, telemedicine, insurance, banking, legal, pharma, defense, and government.

Dograh offers a workflow builder interface where users can visually configure and swap components in their voice agent stack. The platform supports integration with a variety of open-source and proprietary STT and TTS models, including Whisper, Voxtral, Kokoro, and Chatterbox, as well as telephony providers like SIP and Twilio. Users can also bring their own voice clones and model weights. Dograh’s Model Context Protocol (MCP) server allows developers to build, modify, and deploy voice agents directly from agent runtimes or IDEs, with compatibility for tools such as Claude Code, OpenClaw, and Cursor.

Deployment options include self-hosting via cloning the repository and running with Docker Compose under a BSD 2-Clause license, managed cloud hosting by Dograh, or a managed private cloud setup within the user’s own VPC. The self-hosted option is free and designed for organizations seeking maximum data control. The platform also supports combining pre-recorded voice clips with TTS in the same cloned voice, optimizing for latency and cost while enhancing the naturalness of agent interactions.