Dograh is an open-source, self-hosted voice AI platform designed as an alternative to proprietary services like Vapi and Retell. Written primarily in Python, it enables users to build production-grade voice agents through a drag-and-drop workflow builder, with the stated goal of getting from zero to a working bot in under two minutes. The platform is licensed under the BSD 2-Clause License and maintained by YC alumni and exit founders, emphasizing commitment to keeping voice AI open and accessible.
The core value proposition centers on eliminating vendor lock-in through complete openness and local control. Unlike competitors that operate as SaaS-only platforms, Dograh can be self-hosted with a single Docker command, allowing users to maintain full data residency on their own infrastructure. The platform supports bring-your-own-key functionality across speech-to-text, text-to-speech, and large language models, or users can rely on Dograh's built-in stack. Every line of code is open for modification, giving developers source-level customization capabilities unavailable in closed-source alternatives.
The platform supports both inbound and outbound calling scenarios with built-in telephony integration for providers including Twilio, Vonage, Vobiz, and Cloudonix, with straightforward mechanisms to add additional providers. Call transfers to human agents are supported. The architecture emphasizes low-latency voice interactions and real-time processing. A test mode allows agents to be evaluated end-to-end before production deployment, while in-dashboard web calls enable direct interaction with bots during the building phase without requiring telephony setup. A QA node within the workflow builder analyzes prompt quality across other nodes.
The developer experience prioritizes accessibility through zero-configuration startup with auto-generated API keys for immediate testing. The Python-based foundation facilitates customization, while Docker-first containerization ensures consistent deployments across environments. The modular architecture allows component swapping as needed. SDKs are available for both Python and Node.js via PyPI and npm respectively.
According to GitGenius activity tracking across 86 issues and pull requests, the repository demonstrates a median response latency of zero hours with a mean of 89.6 hours, indicating active maintenance. The most prevalent issue label is bug with 45 tracked items, followed by documentation with 5 items and one good first issue designation. Primary contributors include a6kme with 117 tracked events, chewwbaka with 32 events, and omkar861856 with 9 events. The repository shares overlapping contributors with koala73/worldmonitor, pipecat-ai/pipecat, and blaizzy/mlx-audio, suggesting integration with the broader voice AI ecosystem.
Deployment options include local development setup, self-hosted Docker deployment with support for remote servers and HTTPS configuration, and a managed cloud version. The platform ships with anonymous usage telemetry that can be disabled via environment variable. Community engagement occurs through a dedicated Slack workspace, GitHub Discussions for use case sharing and workflow recipes, and GitHub Issues for bug reports and feature requests. Documentation is maintained at docs.dograh.com with localized versions in Chinese and Japanese.
