A full open-source alternative to Voiceflow for phone agents is Dograh, an open-source, self-hostable voice agent platform under BSD-2-Clause that anyone can run on their own servers for free. Voiceflow is hosted, with private cloud hosting on its Enterprise plan, and its platform code stays closed. Call audio stays in-house only when the models are self-hosted too.
Key Takeaways
- Voiceflow's private cloud is an Enterprise option; the platform code stays closed.
- To get the code without a sales contract, self-host Dograh under BSD-2-Clause.
- Call audio stays in-house only when the models run on your servers too.
This post is part of our guide to Voiceflow alternatives. The guide's section "Open-source frameworks: LiveKit Agents and Pipecat" separates a full platform from a framework in a few paragraphs. This post explains how private cloud differs from open source, and what you take on when you run the code yourself, with Dograh, an open-source, self-hostable voice agent platform, as the example.
What Voiceflow offers for private hosting
Voiceflow runs your agent on its own cloud, and it does offer a private option, so the honest comparison starts there. Voiceflow's pricing page says its production platform is "all hosted on Voiceflow". Its security page says the platform is built on AWS and GCP with isolated environments. The model providers it connects, such as OpenAI, Anthropic and Google, are set up with zero data retention, and Voiceflow does not train models on customer data. When you switch on call recordings, they are stored by Voiceflow.
For teams that need more, private cloud hosting is listed on the Enterprise plan, next to single sign-on and custom credit allotments. Voiceflow also lists SOC 2 Type II, ISO/IEC 27001:2022 and GDPR compliance, and supports HIPAA deployments with a BAA arranged through its security team.
All of that is real, and for many support teams it is enough. But in every setup the software stays closed. Voiceflow's public GitHub holds libraries, a command-line tool and demos, while the platform itself stays private. Private cloud runs Voiceflow's software somewhere more private, and the code still stays with Voiceflow.
What an open-source alternative to Voiceflow adds
Open source changes who controls the platform, which private cloud alone leaves with the vendor. Dograh's code is on GitHub under the BSD-2-Clause licence, which lets you use, change and redistribute it, commercially too, as long as the copyright notice stays. In practice that gives you four things a closed private cloud leaves with the vendor. You can install it today, with no sales call. Your security team can read the code it is asked to approve. You can change it, or pay someone to, when a feature is missing. And you can keep running the version you have if the vendor's pricing or direction changes.
That last point is the lock-in question. In a Zapier survey of US enterprise executives, only 6% said they could stop using their main AI vendor without disruption. With open source, the worst case is that you maintain your own copy.
The table sets the private routes against each other, with no prices, since Voiceflow's Enterprise plan is quoted on request.
| Question | Voiceflow Enterprise (private cloud) | Dograh, self-hosted | LiveKit Agents or Pipecat |
|---|---|---|---|
| Can you read and change the code? | No, the platform is closed | Yes, BSD-2-Clause | Yes, Apache-2.0 or BSD-2-Clause |
| How do you get it? | An Enterprise contract with Voiceflow | Install it today with Docker Compose | Install the framework and write the agent in code |
| What comes with it? | Voiceflow's builder, chat, phone and SMS channels, native CRM and helpdesk tools. But the software and its updates stay Voiceflow's | Builder, telephony, Transfer Call, Campaigns, QA and traces. But you choose and run the model layer | Code building blocks for a voice pipeline, with any model you wire in. But your developers write the agent and the product around it |
| Does call audio stay in-house? | It passes through Voiceflow's software and the model providers it connects, which keep no copies | Only when you also host open-weight models | Only when you also host open-weight models |
| Who runs it? | Voiceflow, in a cloud set up for you | Your team, with optional Enterprise support | Your team |
A full platform or a framework
Open source alone does not give you a product. With a framework, your developers build the product. With a full platform, they build the agent. LiveKit Agents and Pipecat are good open-source frameworks. LiveKit Agents runs on the LiveKit media server under Apache-2.0, with more than 14,800 GitHub stars. Pipecat is a Python framework from Daily under BSD-2-Clause, with more than 16,700, and Dograh itself is built on a fork of it. Pipecat is an excellent base, but a production phone agent needs a lot of extra plumbing: phone integrations, call history, quality assurance (QA) on live calls and the call lifecycle.
Dograh ships that plumbing in the same repository, which has more than 5,900 stars. A visual workflow builder draws the call as nodes joined by pathways, each with a written condition the AI reads, so a team used to drawing flows in Voiceflow will find the idea familiar. Telephony connects to your own carrier, such as Twilio, Vonage, Telnyx or Asterisk ARI, so your numbers stay with that carrier.
Transfer Call hands the caller to a person, Campaigns dial outbound lists with retries and calling hours, and a QA node scores each step of a call after it ends. Every call keeps its transcript and recording. Our Dograh vs Voiceflow comparison goes through each feature row by row.
Open Source Alternative to Vapi / Retell
Self-hosted voice agent platform — no per-minute fees
dograh-hq/dograh
Star on GitHub
Keeping call audio in-house
Self-hosting the platform keeps call records on your servers. Keeping the audio there too takes one more step. Dograh is the orchestration layer: it runs the call flow, tools and telephony, and the speech and language models are a separate layer you choose. If that layer is a provider on your own key, the audio still goes to that provider, because bringing your own key (BYOK) moves the contract, not the data. The call audio stays in-house only when the speech and language models run on your own servers too.
The language and speech-to-text models we recommend fit together on one 24 GB GPU: Qwen3.8 27B, quantized to 4-bit, with Parakeet TDT 0.6B v2 beside it at roughly 2 GB. For the voice, Magpie TTS gives fixed voices, and Chatterbox can clone one. Dograh connects to local language models through OpenAI-compatible servers such as Ollama or vLLM, and to local speech models through Speaches. Our tutorial on building a voice agent on fully local open-source models walks through that setup.
Self-hosted Dograh can also run offline when set up this way: local models, calls through your own phone system over Asterisk ARI or through the browser, and telemetry turned off. Telemetry is on by default. Set ENABLE_TELEMETRY=false, and remove POSTHOG_API_KEY from the api service in docker-compose.yaml (or set it to an empty value). Installing and updating still need internet access or a private image registry. The full list of server settings is in Dograh's environment variables reference.
What you take on when you run it yourself
Owning the platform moves operations work to your team, and that is the fair objection to open source. A self-hosted install runs with Docker Compose on a server with at least 8 GB of RAM and 4 vCPUs, with no GPU when your models run through API keys. It runs 10 calls at once by default, a limit you can raise, and your servers and phone carrier set the real ceiling. Self-hosting puts uptime and upgrades in your team's hands, along with monitoring, backups and incident response.
Recordings and transcripts have no retention setting, so deleting old calls is a job for your own storage. Recordings and transcripts keep whatever is said, so the call never asks for a full account, card or Social Security number; use a short check such as the last four digits, and send payment links by text or email from your own system.
For patient calls, self-host Dograh. No BAA with Dograh is needed, because Dograh never sees the data. Your speech and language providers may need their own BAA, and your phone carrier may too. If the servers run in a cloud account such as AWS, that cloud provider needs one as well. Dograh starts every account on its own managed models. To keep call audio away from Dograh, open Models, switch the Organization default to BYOK with your own speech and language providers (or local models), and check that no agent has its own model override. Audit logs are listed on the Enterprise plan.
Self-hosted Dograh has no per-minute platform fee (servers, model keys and phone minutes still cost); Dograh Cloud has a 1¢-a-minute platform fee. If you want the platform without running it, Dograh Cloud handles the security work for you.
When Voiceflow's own hosting is enough
Open source is worth the operations work only if you need what it gives you. Dograh is built for phone calls and also offers a text chat widget for websites. Teams whose main channel is text chat may prefer Voiceflow. Stay too if you want no servers at all, are happy with zero data retention at the model providers, and need the vendor to hold SOC 2 and ISO 27001 reports. Self-hosted Dograh covers that need its own way: it runs inside your own perimeter, so there's no vendor to certify.
If you need the code, your own carrier or call audio that stays in your building, open source is the route. Our guide to migrating from Voiceflow to open-source voice AI shows how to move one number at a time, with Voiceflow kept as the fallback. To start, clone Dograh from GitHub and run it on your own server, or try it on Dograh Cloud.
Join the Dograh Community
Dograh is an OSS alternative to Vapi. Join our Slack community for queries, releases, best practices & community interactions.
Glossary
- Private cloud hosting
- Running a vendor's platform in a cloud set up for one customer, still under the vendor's contract and closed code. Voiceflow lists it on its Enterprise plan.
- BSD-2-Clause licence
- A permissive open-source licence. It lets anyone use, change and redistribute the code, commercially too, as long as the copyright notice stays. Dograh and Pipecat use it.
- Open-weight model
- An AI model whose weights you can download and run on your own servers. Hosting one is the only way to keep call audio from reaching an outside model provider.
- Bring your own key (BYOK)
- Using your own account with a model provider instead of the platform's. You hold the contract and the bill, but the call audio still goes to that provider.
