Kore.ai alternatives are divided into open-source platforms you run, hosted voice platforms with published prices and voice-first managed vendors. Where a team wants to own the platform, Dograh is the open-source choice, run self-hosted or on Dograh Cloud. Vapi, Retell AI and Bland are the hosted voice platforms, and PolyAI and Parloa are the voice-first vendors.
Key Takeaways
- Kore.ai already offers on-prem, so the real choice is open code or closed.
- Dograh is open source and self-hosts with Docker Compose on one server.
- Vapi, Retell AI and Bland publish per-minute rates you can price yourself.
Kore.ai covers service, employee and back-office automation, so teams that only need voice agents often look elsewhere. We put every option in one table, start with open-source Dograh that you run yourself, then cover hosted and voice-first vendors, price a 40,000-minute month, and explain staying, leaving and choosing.
Kore.ai alternatives in one table
Kore.ai sits in the first row, and each option after it answers the same five questions. Kore.ai offers multi-tenant SaaS, a dedicated VPC and on-premises deployment, so a buyer can already keep Kore.ai inside their own walls. The question this table asks is what else you get, and what you give up.
| Option | Open source | Where it runs | How it bills | Scope |
|---|---|---|---|---|
| Kore.ai (baseline) | No | Multi-tenant SaaS, a dedicated VPC or on-prem | By quote, modules metered per session on a 12-month contract | Customer service, employee support and process automation, voice through its own gateway |
| Dograh | Yes, BSD 2-Clause | Your own servers with Docker Compose, on-prem on Enterprise, or Dograh Cloud | Self-hosted, no per-minute platform fee. Dograh Cloud, 1¢/min platform fee. Models and phone extra | Voice agents for phone and browser calls, with a web chat widget |
| Vapi | No | Vapi's cloud | $0.05/min hosting plus models and phone. Pro has a $999/mo minimum for 30 calls at once | Voice agents for developers |
| Retell AI | No | Retell's cloud, with on-prem through Enterprise sales | $0.055/min voice infrastructure and $0.015/min telephony, plus the LLM. $0.07 to $0.31/min all-in | Hosted voice agents |
| Bland | No | Bland's cloud, with on-prem on Enterprise | Start $0.14/min. Build $0.12/min plus $299/mo | Voice agents on Bland's own models |
| PolyAI | No | PolyAI's cloud | Essentials $49/mo for 250 minutes. Scale $299 per location. Enterprise by quote | Voice-first agents, self-serve or built for you |
| Parloa | No | Parloa's cloud, with customer data kept in a chosen region | By quote, per minute for voice | Voice-first agents, with chat on the same platform |
| Cognigy | No | Cognigy SaaS or NiCE CXone. On-prem closed to new customers | By quote, per billable conversation | Voice and digital channels, plus tools for human agents |
| LiveKit Agents and Pipecat | Yes, Apache-2.0 and BSD-2-Clause | Your own servers | Free to self-host | Frameworks your developers build on |
Half the rows can run inside your own walls under some contract, but only three let you read the code, and of those only Dograh comes as a finished product. The other closed rows run your calls on code you cannot inspect, wherever the servers sit.
Dograh: an open voice platform you own end to end
Kore.ai already sells on-prem, so the question is whose code runs there. Dograh's full platform is public in its GitHub repository under BSD 2-Clause, from the builder to the phone connections. You can self-host it with Docker Compose on one server with 8 GB of RAM and 4 vCPUs, run it on Dograh Cloud, or take on-prem and private-cloud deployment with the Enterprise plan.
Agents are not written in a vendor language. In the Agent Builder you describe the agent in plain words, approve the "Your agent requirements" card and let it build, then edit by hand on the canvas when you want. A Dograh MCP server also lets an AI assistant create and edit agents, and an agent exports as JSON that another install can load through Upload Agent Definition.
As the orchestration layer, the platform runs the call flow, tools and telephony, and you choose the speech-to-text, LLM and voice for each agent. Those can be its own models, your provider keys or open-weight models on your own machines, and seven speech-to-speech options are built in. Your own keys move the contract to you while the audio still reaches each provider. Transfers work on Twilio, Telnyx, Plivo, Cloudonix and Asterisk ARI, where Dograh calls the destination first and connects the caller only when someone answers.
The hosting choice sets the bill. Self-hosted Dograh has no per-minute platform fee, but servers, model keys and phone minutes still cost money. Dograh Cloud charges a 1¢-a-minute platform fee and includes 10 concurrent calls, with more on request, and a self-hosted install defaults to 10, which you can raise. Owning it means owning the upkeep too, since the server, the database and the upgrades fall to your team, with community support on open source and custom SLAs on Enterprise.
Open Source Alternative to Vapi / Retell
Self-hosted voice agent platform — no per-minute fees
dograh-hq/dograh
Star on GitHub
Hosted voice platforms with published rates
If the reason to leave is a quote you cannot check, these four let you price the platform before a call with sales. Vapi charges $0.05 a minute for hosting, with models and the phone line on top. Its Pro package sets a $999-a-month minimum, or 10% of hosting if that is higher, and brings 30 calls at once.
Retell AI bills $0.055 a minute for voice infrastructure and $0.015 a minute for telephony, plus the language model, which puts its own all-in range at $0.07 to $0.31 a minute. Twenty calls at once come with every account. Bland's plans start at $0.14 a minute with its own models, and Build lowers that to $0.12 for $299 a month, with 50 calls at once and 2,000 a day.
PolyAI's self-serve Essentials plan is $49 a month for 250 minutes, and its Scale plan is $299 per location. All four keep the code closed and the runtime on their own servers by default, so call recordings and transcripts sit with the vendor.
Voice-first vendors for teams that want it built
Some teams leave Kore.ai for the opposite reason. They want less platform to manage and a vendor that builds and tunes the agent for them. For that buyer, self-hosting is the wrong answer, and a voice-first managed vendor is the right one.

PolyAI Enterprise is quoted by sales and comes with a HIPAA BAA, single sign-on, an uptime SLA and data storage in your region, with PolyAI or a partner building the agent. Parloa prices through its team, per minute for voice, keeps customer data in the region you choose, and lists more than 140 languages. Both stay closed source, and both plug into contact centers such as Genesys, Five9 and Amazon Connect.
LiveKit Agents and Pipecat for teams that build
Developers who want full control and are ready to build the product can start from a framework. LiveKit Agents is Apache-2.0, with its turn-detection models under a separate LiveKit Model License, and Pipecat is a BSD-2-Clause Python framework that Dograh itself builds on. Neither ships the phone setup, builder, call history or QA, so your team writes those. Our comparison of open-source voice AI frameworks covers the trade in detail.
What 40,000 minutes a month costs
Published rates are easiest to judge at your own volume. As an example, take 40,000 call minutes in one month and price only the platform line each vendor bills.
| Option | Platform line for 40,000 minutes | Still to add |
|---|---|---|
| Kore.ai | By quote, modules metered per session | Each module you add |
| Vapi | 40,000 x $0.05 = $2,000 | Speech, language model, voice and phone |
| Retell AI | 40,000 x $0.055 = $2,200, plus 40,000 x $0.015 = $600 telephony, so $2,800 | The language model. Retell's own all-in range gives $2,800 to $12,400 |
| Bland Build | 40,000 x $0.12 + $299 = $5,099 | Phone time |
| Dograh Cloud | 40,000 x $0.01 = $400 | Your models and carrier |
| Dograh self-hosted | No platform fee | Servers, model keys and phone minutes |
| Cognigy, Parloa, PolyAI Enterprise | By quote | Set in the quote |
The lines do not cover the same things. Bland's rate includes its models, while the Vapi, Retell AI and Dograh lines still need speech and language models added. For Kore.ai, with session-based metering, the number to get in writing is how one voice call turns into sessions on your contract. Our post on Kore.ai pricing lists the questions that pin that down.
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 Kore.ai makes sense
Kore.ai does a lot that a voice platform does not try to do. If you use AI for Work for HR and IT requests, or AI for Process for back-office jobs, replacing only the voice agent can leave you running two vendors. Its governance tools, its on-prem and dedicated VPC options and its certification list also suit large regulated buyers.
Staying also makes sense if your contact center runs on Genesys, Five9, NiCE CXone, Amazon Connect or Talkdesk and Kore.ai is already wired into it. Our Kore.ai review covers who it fits. If you want to switch suites, Cognigy is the closest match, though new customers can buy it only as SaaS, and our roundup of Cognigy alternatives covers that move.
What moving off Kore.ai takes
A move off Kore.ai is a rebuild, because agents are written in its own Agent Blueprint Language (ABL). Print or export the agent definitions you have and use them as the spec, then describe each agent again in the new builder.
Call history comes next. Recordings can be fetched per session through Kore.ai's API, so pull the ones you need for testing and audits before the contract ends, and ask Kore.ai what a full export includes. For phone numbers, Kore.ai's voice docs show two routes. Numbers bought through Twilio inside Kore.ai need a port question, while a SIP trunk you brought yourself can simply point at the new platform.
Then move in stages. Send one line or one queue to the new agent and leave the rest on Kore.ai. Test every transfer over a real phone call, and use the QA step to score each step of the rebuilt calls against the old ones before the next line moves.
Picking your Kore.ai alternative
Begin with the reason you are leaving. If it is control, meaning open code, your own servers or on-prem with code you can read, start with Dograh, and add an Enterprise contract when you want on-prem support behind it. If your developers want to build from parts, use LiveKit Agents or Pipecat. If it is price, Vapi, Retell AI and Bland let you run the numbers at your own volume.
If you want a vendor to build and run the voice agent, PolyAI Enterprise and Parloa fit that buyer. And if you need another full suite across chat and voice, look at Cognigy, or stay on Kore.ai. Before any move, run one real call flow on the option you prefer.
Glossary
- Session-based metering
- Billing that counts sessions rather than minutes. Kore.ai meters each AI for Service module per session on a 12-month contract, so ask how a voice call becomes sessions.
- Agent Blueprint Language (ABL)
- Kore.ai's own typed language for defining agents. Agents written in it run only on Kore.ai, so moving means rebuilding them.
- Platform fee
- What a platform bills for running a voice agent's call, kept apart from the models and the phone line. Vapi charges $0.05 a minute for hosting, and Dograh Cloud charges $0.01.
- Orchestration layer
- Everything in a voice agent that runs the call, from call flow and turn-taking to tool calls and the phone connection. The speech and language models are a separate layer you select.

