Dograh Review (2026): Pricing, Features & Honest Verdict

Dograh is an open-source platform for building voice AI agents that handle phone calls, positioned directly against Vapi and Retell. Self-host it free, or use the managed cloud at one cent per minute. Best for: teams in regulated industries that cannot ship call audio to a closed vendor. Price: free self-hosted, cloud from $0.01/min. Rating: 8/10.

What is Dograh?

Dograh is voice agent infrastructure. You build a calling agent in a visual workflow editor, wire it to a telephony provider, and it handles live phone conversations: answering inbound calls, making outbound ones, booking appointments, qualifying leads, chasing payment reminders, and handing off to a human when the conversation needs one. The whole platform is published under a BSD-2 licence on GitHub, where it has collected around 5,200 stars, and the company runs a managed cloud on top of the same code.

The pitch is aimed squarely at the two incumbents. The page title reads “The Open Source Vapi Alternative” and the meta description names Vapi and Retell in the same sentence. What Dograh is selling against them is not features but control. The founders’ own framing on the company page is that the platforms building voice agents on top of large language models “locked the stack behind closed APIs, opaque billing, and forced data flows.” Their customers, they say, work in fintech, healthtech, defence and regulated geographies and cannot afford to leak. That is a real market, it is underserved, and open source is the honest answer to it.

What are Dograh’s key features?

Every box in the pipeline is swappable

The homepage has an interactive diagram of the voice pipeline where you click any stage and see the providers you can drop in. Speech-to-text can be Deepgram, AssemblyAI, Speechmatics, Gladia, Groq or several others. The language model can be OpenAI, Gemini, Azure, AWS, Hugging Face, OpenRouter, Minimax or Sarvam. Telephony runs through SIP with Twilio as the reference integration. Nothing is welded in. This is the practical expression of the open-source pitch: if a provider raises prices or your compliance team rules one out, you change a setting rather than rebuild.

Cascade or speech-to-speech, your choice

Most voice agent platforms commit to one architecture. Dograh supports both. You can run the classic cascade, where speech-to-text feeds a language model which feeds text-to-speech, and tune each stage independently. Or you skip the whole chain and run a single speech-to-speech model that takes audio in and returns audio out, with Gemini 3.1 Flash Live as the example provider. Cascade gives you control and cheaper components; speech-to-speech gives you lower latency and more natural turn-taking. Being able to test both on the same agent without rewriting it is a real advantage, because which one wins depends on your call type and you will not know until you try.

A no-code workflow builder over real infrastructure

The agent itself is built in a visual flow editor rather than in code, which means an operations lead can adjust a call script without filing a ticket. That combination, a no-code surface sitting on infrastructure you can fork, is unusual. Most no-code voice tools are closed, and most open-source voice stacks are libraries you assemble yourself. Dograh is trying to be both, and the workflow builder is the reason a non-engineer can own the agent after an engineer has deployed it.

MCP-native control

Dograh exposes the platform through the Model Context Protocol, so a coding agent or an assistant can drive it directly: create agents, adjust workflows, inspect calls. There is also a plugins repository for Claude Code. This is early and the practical payoff depends on how much of your operation is already agent-driven, but it is a sensible bet on where tooling is heading and it costs nothing to have.

Call recording with automated QA and traces

Every call is recorded, and the platform runs automated quality assurance across recordings rather than leaving you to sample them by hand. Real-time analytics, observability and traces sit alongside, so a failed call is debuggable after the fact instead of being a mystery. For a voice agent in production this is the difference between a system you can operate and a system you hope is working, and it is the sort of unglamorous feature that tends to be missing from young platforms.

How much does Dograh cost?

Dograh’s self-serve cloud costs one cent per minute as a platform fee, and the self-hosted version is free forever. Those are the two prices that matter, and both are published clearly.

On the cloud, the one cent per minute is a platform fee only. Model costs sit on top, and you can bring your own API keys or buy usage through Dograh’s own models. Ten concurrent calls are included, with higher concurrency requiring a conversation. You start with credits from $5 and top up as you go, which is about as low as the barrier to a first production call gets. The self-hosted option is the full platform under BSD-2 with no feature gating: the company is explicit that the open-source release is “the full platform, not a stripped trial.”

Above that, everything is quote-based. A committed-volume tier unlocks better rates on committed minutes plus priority support over email, Slack and WhatsApp. An enterprise tier covers on-premise or private-cloud deployment, data residency, SSO and SAML, audit logs, compliance support and a dedicated account manager with custom SLAs. Neither publishes a figure.

That is the main pricing criticism. There is exactly one committed number on the page. If your volume is modest or you self-host, you never need another one. But if you are the fintech or healthtech buyer this product is explicitly built for, the tier you actually need is the on-premise one, and its price is a conversation. The open-source escape hatch takes most of the sting out of that, since you can run the same software yourself if the quote comes back wrong.

How does Dograh’s pricing compare with Vapi and Retell?

This is where the review gets interesting, because the comparison is not close.

Vapi charges $0.05 per minute as its platform fee. Dograh charges $0.01. That is a fifth of the price for the equivalent line item, and both include ten concurrent calls at the entry level. Vapi then charges $10 per line per month beyond the included ten, and prices HIPAA compliance at $2,000 a month and zero data retention at $1,000 a month as separate add-ons. Both platforms pass model costs through at cost if you bring your own keys.

Retell publishes a range of $0.07 to $0.31 per minute depending on configuration, with $10 in free credits and twenty free concurrent calls to start.

So on the headline platform fee, Dograh is five times cheaper than Vapi and roughly seven times cheaper than the bottom of Retell’s range. The caveat is that a platform fee is not a total cost: models, transcription, voices and telephony all bill separately on every one of these platforms, and those costs dominate for most workloads. A five-times difference on one component of the bill is smaller in practice than it looks. But it is real, it goes the right way, and the compliance features Vapi charges thousands a month for are things you get by running Dograh on your own hardware.

Who is actually behind Dograh?

Worth a short section, because the sources disagree and we think the discrepancy is instructive rather than sinister.

The Product Hunt listing credits Rohan Chaubey, Pritesh Kumar and Sandeep Vemu. Dograh’s own company page names a different set: Pritesh Kumar and Abhishek Kumar as co-founders, with Sabiha Khan as founding engineer. Only Pritesh Kumar appears on both. Rohan Chaubey is a prolific Product Hunt hunter who appears on launches from unrelated companies, so his presence on the listing is almost certainly a hunter credit rather than a founder credit.

We are treating the vendor’s own about page as authoritative: this is a three-person team, two co-founders and a founding engineer. That is small for infrastructure that fintech and healthtech buyers are meant to run in production, and key-person risk belongs in your evaluation. The mitigation is the licence. If the team moves on, the code is BSD-2 on GitHub with a few thousand stars behind it, and you are not left with a dead SaaS account.

Who is Dograh best for?

Use Dograh if you are building phone agents in an industry where the data cannot leave your perimeter. That is the case it was built for and the case where nothing else on the market is a clean fit: running Vapi with HIPAA turned on costs $2,000 a month before a single call, while running Dograh on your own infrastructure costs whatever your servers cost. Use it too if you are cost-sensitive at volume, since a five-times lower platform fee compounds, or if you have been burned by a closed vendor changing pricing under you.

Skip Dograh if you want a managed product with a support contract and no infrastructure work. The cloud is real, but the ten-concurrent-call self-serve ceiling means any serious deployment involves a sales conversation, and a three-person team is not going to give you the support surface an incumbent will. Skip it also if you need a specific compliance certification on the vendor’s side rather than on yours; self-hosting moves the compliance burden to you, which is the point but is not always what a buyer wants.

Best Dograh alternatives

Vapi is the incumbent Dograh names. Well funded, recently raised a $50M Series B, and the most mature developer experience in the category. $0.05 per minute platform fee, ten concurrent lines included then $10 per line per month, HIPAA at $2,000 a month and zero data retention at $1,000 a month. Pick it if you want the safest managed option and the budget is not the constraint.

Retell AI publishes $0.07 to $0.31 per minute depending on setup, starts at $0 with $10 in free credits and twenty free concurrent calls, and makes a point of not requiring an annual contract. The most generous free trial of the three and the easiest way to test whether voice agents work for your use case at all.

Pipecat is the other open-source route, an orchestration framework rather than a product. More assembly required and no managed cloud or no-code builder, but total control and a large community. Choose it if you want to build the platform rather than adopt one.

Final verdict: is Dograh worth it?

Yes, particularly if the words “data residency” appear anywhere in your requirements. Dograh is doing the thing the voice AI category needed someone to do: publishing the whole stack under a permissive licence, charging a fifth of what the market leader charges for the managed version, and letting you swap every component instead of locking you to whichever transcription vendor the platform struck a deal with. The engineering looks serious, the observability and automated QA suggest people who have operated this in production, and supporting both cascade and speech-to-speech architectures is a real technical answer rather than a marketing bullet.

The reservations are about company stage, not product quality. Three people are building infrastructure meant for regulated industries, only one price on the pricing page is committed, and the self-serve tier caps at ten concurrent calls, so the plan you actually need is behind a sales conversation. None of that is disqualifying, and the BSD-2 licence is a real hedge against all of it. Start self-hosted, prove the agent works on your call flows, and only then decide whether you want the managed version or your own servers.