Open-source Leaping AI alternatives are divided into platforms that ship ready to take calls and frameworks you build on, with hosted per-minute services as a third route. Dograh is the open-source AI voice agent platform we build, and you can self-host it or run it on Dograh Cloud. LiveKit Agents and Pipecat are the frameworks.
Key Takeaways
- A Leaping quote is one number. An open stack shows every cost line.
- Self-hosted Dograh keeps recordings and transcripts on storage you run.
- Self-hosting means your team owns the server, the database and the upgrades.
Leaping AI quotes each deal and runs your calls on its cloud. This post is for teams that want to run the code. We price a month of lead calls line by line, show where call data lives, rebuild a home-improvement lead flow in Dograh and list self-hosting's real work.
Open-source Leaping AI alternatives in one table
Going open source is a different purchase from Leaping's. You get code you can run and change, and your team takes on the running. Leaping prices by quote and provides a cloud phone system, with European data kept inside EU borders.
| Option | License | Where calls run | What you pay | Ready on day one | Call data |
|---|---|---|---|---|---|
| Leaping AI (baseline) | Closed | Leaping's cloud, with provided numbers or your own phone company | A quote | Yes, Leaping builds the agent | On Leaping's cloud, with European data kept in the EU |
| Dograh | BSD 2-Clause | Your own servers, on-prem on Enterprise, or Dograh Cloud | Self-hosted, no per-minute platform fee. Dograh Cloud, 1¢/min platform fee. Models and phone at their own rates | Yes, with the builder, Campaigns, transfers, QA and traces included | In your own database and storage when self-hosted |
| LiveKit Agents | Apache-2.0, turn-detection models under the LiveKit Model License | Your own servers, or LiveKit Cloud | Free to self-host. LiveKit Cloud free with 1,000 agent minutes, then $50/mo for 5,000 | No, your team builds the product around it | Wherever your team builds it to go |
| Pipecat | BSD-2-Clause | Your own servers, or Pipecat Cloud | Free to self-host. Pipecat Cloud from $0.01/min per active agent | No, your team builds the product around it | Wherever your team builds it to go |
Only one open row arrives ready to take a lead call. The frameworks are sound foundations, and a team that picks one should plan for the product work between a working demo and a production phone line.
Dograh: open code you run yourself
Dograh is open source under BSD 2-Clause, and the repository holds the whole platform, from the call-flow builder to the phone connections. A permissive license means the code you deploy today stays yours to run whatever the company behind it does later. One Docker Compose file starts the stack, with Postgres for runs and transcripts, MinIO for recordings, Redis, nginx, a TURN server, the API and the UI.
Dograh runs the call flow, the tools and the phone connection as the orchestration layer, and the speech and language models sit in a separate layer you pick. Choose Dograh's models, your own provider keys or open-weight models on your hardware. Only the last keeps the caller's voice inside your network, since a provider key moves the contract to you while the audio still travels to that provider.
Running it on your own servers carries no per-minute platform fee, while servers, model keys and phone minutes still cost money. If you would rather skip the server, Dograh Cloud charges a 1¢-a-minute platform fee. On-prem deployment inside your own data perimeter comes with the Enterprise plan.
The work that comes with self-hosting is yours to plan for. Your team keeps the server running, looks after the Postgres database and its backups, and applies each update to a newer image. Help for the open-source edition comes from the community in Dograh's Slack, and the Enterprise plan adds custom SLAs.
The server needs at least 8 GB of RAM and 4 vCPUs, and no GPU unless you also host the models on it. Our post on the open-source alternative to Vapi explains how the platform and the models split between a plain server and GPUs. A self-hosted install handles 10 calls at once by default, a limit you can raise, and your servers and carrier set the real ceiling. On Dograh Cloud, pay as you go includes 10, and more are available on request.
Open Source Alternative to Vapi / Retell
Self-hosted voice agent platform — no per-minute fees
dograh-hq/dograh
Star on GitHub
A quote against a bill you build line by line
Leaping gives you one number once the sales call is done. An open stack gives you five lines you can price tonight. Take an example of 2,000 outbound lead calls of three minutes each, 6,000 minutes in one month, and price each line at its vendor's public rate.
| Line | Public rate | 6,000 minutes |
|---|---|---|
| Platform, self-hosted Dograh | No per-minute platform fee | $0 |
| Platform, Dograh Cloud instead | 1¢/min | 6,000 x $0.01 = $60 |
| Server for self-hosting | At least 8 GB RAM and 4 vCPUs, no GPU | Your hosting provider's monthly price |
| Speech-to-text, Deepgram Nova-3 | $0.0077/min | 6,000 x $0.0077 = $46.20 |
| Phone, Twilio US local outbound | $0.0140/min plus $0.0044/min for Media Streams, and $1.15/mo per local number | 6,000 x $0.0184 = $110.40, plus the number |
| Language model and voice | Your provider's rate | Depends on the models you pick |
| Leaping AI | By quote | Set in the quote |
Deepgram charges $0.0077 a minute for Nova-3 streaming, and Twilio charges $0.0140 a minute for a US local outbound call. The model and voice lines are left open because they swing more than any other with your choice, from a small open model on your own GPU to a large hosted one.
Every line can be swapped on its own. If a speech-to-text provider raises its price, you change one setting and keep the rest of the stack. A single quote cannot show you that, and our breakdown of Twilio voice pricing per minute covers the inbound and toll-free rates too.
Where your calls and data live
Leaping keeps calls on its cloud phone system and says its European data is hosted inside EU borders. For many buyers that is enough. Others need the recordings and transcripts in a system they control, and that is where an open platform changes things.
A self-hosted Dograh install writes call recordings to MinIO object storage and keeps each run and its transcript in Postgres, both on your own servers. Completed calls download as a CSV report, and a Webhook node can push each call's results, recording link and transcript link to your CRM the moment it ends.
The agent travels too. Download Workflow in the editor's menu saves the agent as a JSON file, and Upload Agent Definition loads it into another Dograh install. Dograh keeps recordings and transcripts until your team deletes them, so retention follows your own policy.
Rebuilding a home-improvement lead flow
Leaping's product pages show the flow most of its buyers run. A web lead arrives, the agent calls back fast, qualifies the homeowner and books an estimate, and anything unfinished gets a retry or a text. Each of those steps has a Dograh part you set up yourself.
The callback starts from your lead form or CRM. An API Trigger lets that system start a call with the lead's details the moment the form is saved, so how fast the phone rings depends on your system and your carrier.
Lists of older leads go into Campaigns, where you set calling hours for each weekday in one time zone per campaign and choose how many automatic retries busy and unanswered calls get, from 0 to 10. Leads whose calls went to voicemail are dialed again with Redial Campaign after the campaign finishes. Our post on calling hours and retries for AI campaigns covers the settings and the rules behind them.
Texts and CRM updates go through tools. Leaping sells SMS agents, and in Dograh the same job is an HTTP API tool that asks your own SMS provider to send the estimate confirmation during the call. CRM writes work the same way, and Pre-Call Data Fetch can pull the lead's record before the greeting plays.

When the homeowner asks for a person, Dograh calls the estimator first and connects the homeowner only when someone answers, on Twilio, Telnyx, Plivo, Cloudonix or Asterisk ARI. An unanswered, busy or failed transfer comes back to the agent, which keeps talking with the homeowner.
Open frameworks: LiveKit Agents and Pipecat
LiveKit Agents and Pipecat are the other open choices, and both are frameworks for developers. The LiveKit Agents repository is Apache-2.0, with its turn-detection models under a separate LiveKit Model License. Pipecat is BSD-2-Clause Python, and Dograh is built on a fork of it. To match Leaping's product on either, your team writes the phone connection, the lead callbacks, the call history and the QA. Our comparison of open-source voice AI frameworks goes through each.
Join the Dograh Community
Dograh is an OSS alternative to Vapi. Join our Slack community for queries, releases, best practices & community interactions.
When staying on Leaping AI makes sense
Some teams are better served by staying. If nobody on staff wants to run a server, apply upgrades or write the CRM requests, Leaping's done-for-you model takes that work away. Its agent already follows home-improvement playbooks, texts from the same system and connects to HubSpot and Lead Perfection.
Staying also makes sense when the quote works for you and European hosting covers your data rules. Move when you need code you can audit or call data on your own machines.
What moving off Leaping AI takes
Plan the move as a short project with a clear order. First ask Leaping for what you can take with you, such as the scripts, routing rules, recordings and transcripts. Then stand up Dograh on a server, connect your carrier and give the install its own domain with SSL.
Next, rebuild the agent. A self-hosted install with a Dograh Service Key can use the Agent Builder, where you describe the agent, answer a few questions and approve the "Your agent requirements" card before it builds, and "Test agent" saves it and publishes version 1. Without a key, you build the same flow on the canvas.
Then run both. Point one lead source at the new agent, leave Leaping on the rest and compare booked estimates for a few weeks. Test every transfer over a real phone call, since browser test calls cannot transfer, and move the next source only once the first holds up.
Picking your Leaping AI alternative
If open source and your own servers are why you are looking, Dograh comes first, as the one open option that takes calls on day one, with on-prem on Enterprise when you want a contract behind it. If your developers want to build their own product, start from LiveKit Agents or Pipecat. If all you need is a visible price and hosting elsewhere is fine, our roundup of Leaping AI alternatives covers the hosted platforms with published rates. And if a done-for-you vendor still suits you best, stay with Leaping.
Before you commit, build one lead flow and run it on a test number for a week. The install, the bill and the call data are all in front of you from the first call.
Glossary
- Permissive license
- An open-source license such as BSD 2-Clause or Apache-2.0 that lets you run, change and redistribute the code with few conditions. Code you deploy under one stays yours to run if the project changes direction.
- API Trigger
- A Dograh endpoint that lets another system, such as a lead form or CRM, start an outbound call with the contact's details. It is how a fast callback to a new lead is set up.
- Media Streams
- Twilio's feature for streaming live call audio to an outside server, billed at $0.0044 a minute on top of the call rate. A voice agent on Twilio uses it to hear and answer the caller.
- Redial Campaign
- A Dograh option that creates a new campaign to dial contacts again whose last call ended in voicemail, no answer or busy, whichever you tick. Contacts already reached on a retry are skipped.

