The open-source alternatives to Vapi are Pipecat, LiveKit Agents, TEN Framework and Bolna. Pipecat (BSD-2-Clause) and LiveKit Agents (Apache-2.0) are the two with plain open licences, Python frameworks that connect a speech-to-text model, a language model and a text-to-speech model into a live conversation over WebRTC or a phone line. TEN Framework is source available: Apache-2.0 with conditions from Agora that forbid competing deployments. Bolna is MIT and its maintainers say they are looking for help. A team switches when it wants to own the pipeline and choose every model in it.
3 of 4 carry an OSI licence. 4 of 4 run on your own machines. Figures read from GitHub on October 26, 2026.
One project per need, with the reason from its own README and the one thing to know before you choose it.
BSD-2-Clause, maintained by Daily. Over twenty speech-to-text and language providers, over thirty voices including local Whisper, Ollama, Kokoro and Piper, and telephony serialisers for Twilio, Telnyx, Plivo, Vonage, Exotel and Genesys.
Know firstInterruption handling is not described beyond voice activity detection, and every provider is its own account and bill.
Apache-2.0. Semantic turn detection, native MCP tool support, a built-in test framework, and a media server plus SIP service that are open source too, so the entire path can run on your machines.
Know firstBeyond console testing it needs a LiveKit server, cloud or self-hosted, and the turn-detection models carry a separate model licence.
A visual designer, a SIP extension with a Twilio example, turn detection, and avatar extensions, run with Docker Compose.
Know firstSource available, not open source: Apache-2.0 plus Agora conditions that forbid competing deployments, and the quick start requires an Agora account.
MIT. Define the agent in a file, start four containers, and take calls on Twilio or Plivo with language models routed through LiteLLM.
Know firstThe README says the maintainers are looking for help and that maintenance has fluctuated with their time.
The facts that decide most choices. Each project's page carries the full panel.
| Project | Best for | Openness | Licence | Runs as | Own machines | Stars |
|---|---|---|---|---|---|---|
| Pipecat | The widest provider and carrier list | OSI | BSD-2-Clause | pip, CLI | Yes | 16,834 |
| LiveKit Agents | The whole stack on your own servers | OSI | Apache-2.0 | pip (Python), npm (the separate AgentsJS repository) | Yes | 14,943 |
| TEN Framework | A visual designer, if Agora fits | Source available | See LICENSE | Docker Compose, GitHub Codespaces | Yes | 11,201 |
| Bolna | A phone agent from a JSON file | OSI | MIT | Docker Compose, Python library | Yes | 803 |
Licence and star figures read from GitHub on October 26, 2026. Stars are shown as a dated fact and were not used to rank this page.
A Python framework under BSD-2-Clause, maintained by Daily, for real-time voice and multimodal agents, with the widest provider list on this site and telephony serialisers for Twilio, Telnyx, Plivo, Vonage, Exotel and Genesys.
BSD-2-Clause, maintained by Daily. Over twenty speech-to-text and language providers, over thirty voices including local Whisper, Ollama, Kokoro and Piper, and telephony serialisers for Twilio, Telnyx, Plivo, Vonage, Exotel and Genesys.
Know firstInterruption handling is not described beyond voice activity detection, and every provider is its own account and bill.
An Apache-2.0 framework for real-time voice agents that connects any speech, language and voice provider, with semantic turn detection, native MCP tool support, a test framework, and an open-source media and SIP stack it can run on entirely.
Apache-2.0. Semantic turn detection, native MCP tool support, a built-in test framework, and a media server plus SIP service that are open source too, so the entire path can run on your machines.
Know firstBeyond console testing it needs a LiveKit server, cloud or self-hosted, and the turn-detection models carry a separate model licence.
A real-time conversational AI framework from Agora with a visual designer, a SIP extension and turn detection, released under Apache-2.0 plus conditions that forbid competing deployments, which makes it source available rather than open source.
A visual designer, a SIP extension with a Twilio example, turn detection, and avatar extensions, run with Docker Compose.
Know firstSource available, not open source: Apache-2.0 plus Agora conditions that forbid competing deployments, and the quick start requires an Agora account.
An MIT voice-agent platform that defines an agent in a JSON file and orchestrates speech to text, a language model and text to speech over websockets, with Twilio and Plivo for phone calls, whose maintainers say they are looking for help.
MIT. Define the agent in a file, start four containers, and take calls on Twilio or Plivo with language models routed through LiteLLM.
Know firstThe README says the maintainers are looking for help and that maintenance has fluctuated with their time.
The questions that settle it, in the order they usually come up.
LiveKit Agents, whose server and SIP service are open source. Pipecat runs locally with eight transports. TEN's quick start depends on Agora.
Pipecat has serialisers for Twilio, Telnyx, Plivo, Vonage, Exotel and Genesys. Bolna runs on Twilio or Plivo. LiveKit uses its SIP service. TEN has a SIP extension with a Twilio example.
LiveKit Agents describes semantic turn detection with a transformer model. TEN has a turn-detection extension. Pipecat lists voice activity detection through Silero.
Pipecat (BSD-2-Clause), LiveKit Agents (Apache-2.0) and Bolna (MIT). TEN is source available under Agora's added conditions.
Vapi, in the vendor's own words, is "voice agents for builders" that lets you "build, deploy, and improve voice agents as fast as you ship", by its own description. vapi.ai. The cases below are where that still wins.
Stars, licence, last commit and last release are read from the GitHub API by a script every week and carry the date they were read. Nobody types them.
What each project does is written from its own README and LICENSE, read in full on the date shown, and phrased as the project's claim. The sources are listed at the end.
That a project is faster, better or cheaper than Vapi. A price for anyone. A user count. A roadmap. If a fact is not on this page, it was not in the source.
Pipecat (BSD-2-Clause, maintained by Daily) if you want the widest list of speech, language and voice providers and telephony serialisers for Twilio, Telnyx, Plivo, Vonage, Exotel and Genesys. LiveKit Agents (Apache-2.0) if you want to run the whole stack including the media server on your own machines, with semantic turn detection and a built-in test framework. Both are Python; LiveKit also has a Node.js version.
Yes. Pipecat has serialisers for Twilio, Telnyx, Plivo, Vonage, Exotel and Genesys and its CLI scaffolds a phone bot. LiveKit Agents works with LiveKit's telephony stack, which includes an open-source SIP service. TEN has a SIP extension and a Twilio example. Bolna supports Twilio and Plivo, with Exotel and Vonage marked coming soon.
Pipecat's README lists over twenty speech-to-text providers, over twenty language model providers, and over thirty text-to-speech providers, including local options like Whisper, Ollama, Kokoro and Piper. LiveKit Agents describes a mix-and-match ecosystem and names OpenAI, Deepgram, Cartesia and Google in its samples. TEN's quick start requires OpenAI, Deepgram and ElevenLabs keys. Bolna routes language models through LiteLLM.
LiveKit Agents says you can run the entire stack on your own servers, including the LiveKit media server, or use LiveKit Cloud. Pipecat runs locally and lists eight transports, of which Daily is one. TEN's documented quick start depends on Agora's real-time service. Bolna runs in Docker with a Twilio or Plivo account and ngrok for calls.
Pipecat (BSD-2-Clause), LiveKit Agents (Apache-2.0) and Bolna (MIT) are, with LiveKit's turn-detection models under a separate model licence. TEN Framework is source available: its licence is Apache-2.0 plus conditions from Agora that forbid hosting on end-user devices and any deployment that competes with Agora's offerings. Only its packages directory is plain Apache-2.0.
LiveKit Agents describes semantic turn detection with a transformer model to reduce interruptions. TEN's ecosystem includes a turn-detection extension for full-duplex conversation. Pipecat lists voice activity detection through Silero and noise reduction providers; its README does not describe interruption handling. Bolna's README does not mention it.
Reveneau sets up the replacement on your own servers, connects it to your product, and builds the software around it.