Price terms as published

Each line is the vendor’s own wording, split from the price capture checked on 2026-09-23. Nothing here is rounded, converted or re-worded — the full sentences stay in the review below. Lines that carry only a price continue the plan named on the line above, in the vendor’s own order.

2 price terms for Dograh
#Plan / termChecked
1Free (self-host, BSD-2-Clause)2026-09-23
2cloud from $1/min platform fee, credits from $52026-09-23

One-liner: Dograh is a self-hostable, open-source voice AI platform — “the open-source, self-hostable alternative to Vapi & Retell” — where you build voice agents in a drag-and-drop workflow builder, wire up your own LLM/STT/TTS providers, connect a phone line, and launch. Unlike the closed SaaS competitors, every line is yours to modify, your data stays on your infrastructure, and there is a hosted cloud if you would rather not operate it.

Verified facts (GitHub API + raw README + cloud pricing page, 2026-09-23)

Attribute Value Source
GitHub stars 5,715 (1,435 forks, 44 watchers) dograh-hq/dograh via GitHub API, 2026-09-23
Open issues vs PRs API combined 87 — same-day search-API split: 51 issues + 36 PRs GitHub API + search API
License BSD 2-Clause GitHub API + raw LICENSE (Copyright (c) 2025, Zansat Technologies Private Limited)
Language / stack Python backend, modular provider architecture README “Developer Experience” section
Created / last push 2025-09-09 / 2026-09-23 GitHub API
Latest release dograh-v1.47.0, published 2026-09-15 GitHub Releases API
Pricing (self-host) Free, BSD-2-Clause, no feature gating README comparison table
Pricing (cloud) Self-serve $1/minute platform fee, credits from $5, 10 concurrent calls included dograh.com/pricing
BYOK Any LLM / STT / TTS / telephony provider — or Dograh’s own stack with auto-generated keys README “Get Started”
Telephony Twilio, Vonage, Telnyx, Plivo, Vobiz, Cloudonix, Asterisk ARI; human handoff with call transfer README “Voice & Telephony”
Self-host One Docker command (docker-compose.yaml + start_docker.sh) README “Get Started”
SDKs Python (dograh-sdk on PyPI), Node (@dograh/sdk on npm) README “SDKs”
Privacy Anonymous telemetry collected by default; opt out with ENABLE_TELEMETRY=false README “Get Started” notes

What the positioning actually claims

The README’s comparison table is the clearest statement of intent — Dograh scores “None” on vendor lock-in and “Your infra, your rules” on data residency against “Full” / “Their cloud” for both Vapi and Retell. The workflow builder is deliberately n8n-shaped: nodes for actions, tools, transitions, and end-call outcomes, plus a QA node, knowledge bases, webhooks, embeds, and tool calling. Testing ships with Test Audio (speak to it in the browser) and Test Chat (fast prompt iteration), and outbound calling is available through the SDKs. The MCP server is the piece aimed at agent users: connect Codex, Claude Code, or Cursor and the assistant can inspect existing agents, search the docs, fetch node schemas, create workflows, and save draft edits from natural language — the repo even ships a dograh-plugins setup skill that lets your coding agent run the whole install, detect your OS, and pick the deploy path.

The catches

  1. It is a phone-operations tool, not a creator tool. The use cases in the docs are inbound and outbound calling: answer calls, book appointments, qualify leads, run payment reminders. If you came here from a content-creation angle, this is a workflow-builder plus telephony, not a video or audio creation tool — the site’s own positioning is “voice agents”, not “voice content”.
  2. “Free” is only the self-host path. The comparison table’s “Free (self-host)” is accurate but incomplete: the cloud is $1/minute platform fee plus credits from $5, and the 10 included concurrent calls is a hard ceiling until you contact sales. A busy calling operation on cloud Dograh is a per-minute bill that will look like Vapi’s, minus the proprietary license.
  3. You are the operator. Self-hosting means you own the Docker host, the HTTPS setup, and the telephony credentials — the README points you at the remote-server deployment guide rather than doing it for you. Telephony costs (Twilio and friends) are separate line items that this page cannot price.
  4. Release cadence is slower than the commit cadence. The last push was the day of this check, but the newest release is dograh-v1.47.0 from 2026-09-15 — eight days behind, against a 51-issue / 36-PR open queue. v1.47.0 is also the sixth release in the last three months (v1.44.0 → v1.47.0, 07-25 → 09-15), so the rhythm is real, just not daily.
  5. Community badges are the team’s own words. “#1 Product of the Day on Product Hunt” (also #1 of the Week and #3 of the Month) and the “NVIDIA Inception Program” badge come from Dograh’s own site and README, not from an independent source verified for this page.

Who should use it

Teams running or prototyping voice agents on real phone lines who want to keep the stack inside their own perimeter — and who have someone comfortable with Docker and Twilio. Agent users get an unusual bonus: the MCP server means the agent workflow itself can be authored by Claude Code or Codex instead of dragging nodes. If your voice need is dictation or text-to-speech rather than telephony, the local voice tools already covered here are the better fit.

Repository statistics, license text, release history, and the issue/PR split verified via the GitHub API and search API on 2026-09-23; cloud pricing quoted from dograh.com/pricing the same day. Dograh was not installed or run for this page; no hands-on claims are made.

Sources