Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL

dograh-hq/dograh

AI Engineering

Self-hosted voice AI platform for building, testing and deploying phone agents with visual workflows, MCP, and bring-your-own LLM/STT/TTS.

6,198
1,597
Python
BSD-2-Clause
1 year old

What it is

Dograh is an open-source, self-hostable voice AI platform positioned as an alternative to Vapi and Retell. It combines a visual workflow builder (start nodes, agent nodes, global instructions, tools, transitions, end-call outcomes), an in-browser test panel with both voice and chat modes, and telephony integrations for inbound and outbound calls. It is aimed at teams that need production voice agents but want to run the stack on their own infrastructure and pick their own model providers.

The platform is written in Python and ships with its own default LLM/TTS/STT stack and auto-generated keys, so a first agent can be created without external API credentials. You can later attach your own LLM, TTS, STT, and telephony providers (Twilio, Vonage, Telnyx, Plivo, Vobiz, Cloudonix, Asterisk ARI). Storage can use bundled MinIO or any S3-compatible service. A built-in MCP server lets coding agents such as Claude Code, Codex, or Cursor inspect, create, and edit workflows inside a workspace.

Quick start

The README's one-command path downloads a Compose file and a start script, then runs it:

curl -o docker-compose.yaml https://raw.githubusercontent.com/dograh-hq/dograh/main/docker-compose.yaml && curl -o start_docker.sh https://raw.githubusercontent.com/dograh-hq/dograh/main/scripts/start_docker.sh && chmod +x start_docker.sh && ./start_docker.sh

Requirements: Docker on the machine, and roughly 2–3 minutes for the first image pull. The script starts a public *.trycloudflare.com tunnel for inbound telephony webhooks and enables signup by default. To opt out of anonymous usage data collection, set ENABLE_TELEMETRY=false before running; after creating your first account, set ENABLE_SIGNUP=false in .env and re-run ./start_docker.sh (a plain docker compose restart does not apply .env changes).

After startup, open http://localhost:3010, choose Inbound or Outbound, name the bot, describe the use case in 5–10 words, and click Test Agent. Then use Test Audio to speak with the agent in the browser or Test Chat to iterate in text; in chat mode you can edit or replay user turns and Dograh regenerates replies and node transitions from that point. For remote servers with HTTPS, the docs point to a separate Docker deployment path.

Constraints

  • Self-hosting requires Docker; the default local setup exposes a public Cloudflare quick tunnel, which has data-exposure and availability implications for a machine you may not intend to publish.
  • The default quickstart is not locked down: signup stays open until you explicitly disable it in .env.
  • Telephony features depend on a supported provider and the corresponding credentials; human handoff/call transfer is only available on providers that support it.
  • Anonymous telemetry is on unless disabled via ENABLE_TELEMETRY.
  • Interactive turn latency depends on the models chosen; a maintainer cites roughly 500–600 ms when self-hosting models on GPUs, around 800–1000 ms with Gemini-2.5-flash, and higher with larger reasoning models, which is a setup-specific estimate rather than a guaranteed figure.
  • License is BSD 2-Clause.

HotRepo's take

Dograh's own README makes its positioning explicit, and the HN discussion shows the framing needed refinement: a commenter objected that “We open sourced Vapi” sounded like Vapi itself had been released, and the maintainer clarified that it is an independent open-source alternative, not a clone thread. That clarification is consistent with what the repo actually ships—workflow builder, telephony plumbing, SDKs, and MCP—rather than a repackaging of someone else's code.

The project has a real self-hosting story rather than a landing page: Docker quickstart, remote-server docs, Python and Node SDKs, and a plugin for coding agents. Its recorded momentum is also visible on HotRepo: it reached 1,477 stars on May 17, 2026, 2,006 on May 18, and 6,164 by November 17, 2026, when it appeared at position 9 on the Python daily list. Those snapshots are not continuous, so they support the fact of growth, not a trend claim. For readers evaluating voice-agent infrastructure, the practical differentiator is the self-hosted, provider-agnostic deployment; the caveat is that the default local setup opens a public tunnel and signup, which should be closed before anything beyond experimentation.

Written by AI from the project's README, updated . The project's own docs are the source of truth.

Hacker News

Show HN: Dograh – an OSS Vapi alternative to quickly build and test voice agents
▲ 16 · 10 comments · December 8, 2025
Show HN: We open sourced Vapi – UI included
▲ 8 · 7 comments · March 12, 2026

You might also like

  1. rasonyang/ai-native-callcenterSelf-hosted AI call center software that lets a speech-to-speech model answer every call and hands callers to human agents without dropping the line.Go515
  2. vstorm-co/agenticosSelf-hosted platform for teams to build, publish and govern AI agents with shared chat, RAG, tools and budgets.Python345
  3. elie222/rakazoA self-hosted multi-channel AI agent platform for persistent bots with sandboxed computers, aimed at people who want to own the stack.TypeScript5,573
  4. davidondrej/superbotSelf-hosted multi-agent workspace where named AI bots share one Linux VPS and run on coding-agent CLIs.Rust102
  5. 2SinceJuly/SkyDotSelf-hosted multi-user agent workbench for creating named AI companions with shared group chats, scheduled tasks, skills, MCP, and memory.TypeScript183