Brisvox vs LiveKit Agents: a voice agent engine without a media server
Brisvox runs without a media server: WebSocket, Twilio and Telnyx are built in, and WebRTC is an optional cargo feature. LiveKit Agents needs a LiveKit server, Cloud or self-hosted, because the agent joins a room. LiveKit Agents has not been measured against Brisvox, so this page makes no speed or density claim against it.
LiveKit facts checked on against LiveKit's own documentation.
Last updated:
Has Brisvox been measured against LiveKit Agents?
No. LiveKit Agents was not part of the capacity or latency runs. Nothing on this site says that Brisvox is faster, lower in latency or denser than LiveKit Agents. Everything below is what the two projects document.
For orientation only, LiveKit's own deployment guide gives 4 cores and 8 GB for 10 to 25 concurrent jobs, and reports a vendor test of 30 agents peaking at about 3.8 cores and 2.8 GB (LiveKit: deployments). Those are LiveKit's figures on LiveKit's setup. They are not comparable with the Brisvox benchmark, which used different providers, models and limits.
Does LiveKit Agents need a LiveKit server?
Yes. The agent joins a LiveKit room as a participant, so a LiveKit server is required, either LiveKit Cloud
or one you host (livekit/agents). Brisvox takes the call itself: a phone carrier's media stream,
a browser over WebRTC or a plain WebSocket connects straight to brisvox-server.
How do the documented features compare?
| Documented, | Brisvox | LiveKit Agents |
|---|---|---|
| Language | Rust engine; your agent in Python or TypeScript (SDKs on PyPI and npm) or in Rust | Python (livekit-agents 1.8.7); Node.js in a separate repository livekit-agents on PyPIlivekit/agents-js |
| License | MIT or Apache-2.0 | Apache-2.0; the turn-detector text model and v1-mini use the LiveKit Model License livekit/agentsLiveKit: turn detector |
| Runtime model | One process, one isolated async session per call, models on a dedicated thread pool | One OS process per job, “each in its own process for isolation” LiveKit: job lifecycle |
| Needs a media server | No: WebSocket, Twilio and Telnyx built in; WebRTC and SIP are cargo features | Yes: a LiveKit server (Cloud or self-hosted), the agent joins a room livekit/agents |
| Telephony | Twilio and Telnyx media streams, SIP trunk listener, DTMF, transfer | LiveKit SIP service: Twilio, Telnyx, Plivo and others, DTMF, cold and warm transfer LiveKit: telephony |
| Turn detection | Smart Turn v3.2 (ONNX, on device) with an adaptive policy | Own turn-detector models: a text model (396 MB on disk, about 50 to 160 ms per turn on CPU) and the audio model v1-mini (size and cost not stated) LiveKit: turn detector |
| Turkish | First class, with its own quality notes | Listed by both turn detectors LiveKit: turn detector |
| Interruptions | Turn epochs; the context keeps only the text actually played | Adaptive mode separates interruptions from backchannel and resumes after a false one LiveKit: turns |
| Tools | read_only and side_effect kinds, idempotency keys, never speculative; written in Rust, Python or TypeScript | Function tools, async tools, handoff, MCP (Python), provider-side tools; parallel calls not stated on the page read LiveKit: tools |
| Speech-to-speech | OpenAI Realtime, Gemini Live | Many realtime models LiveKit: realtime models |
| Overload handling | Admission control and graceful drain built in | Default load function (5 s average CPU, threshold 0.7) LiveKit: deploymentsLiveKit: server options |
| Hosted cloud | None, self-host only | Optional LiveKit Cloud livekit/agents |
| Vendor sizing guidance | None; see the measured rows | 4 cores / 8 GB for 10 to 25 jobs; a vendor test of 30 agents peaked at about 3.8 cores and 2.8 GB LiveKit: deployments |
How does telephony compare?
Brisvox has Twilio and Telnyx media streams built in, a SIP trunk listener as a cargo feature, DTMF and call transfer. LiveKit Agents uses the LiveKit SIP service, with Twilio, Telnyx, Plivo and other trunks, DTMF, and cold and warm transfer; when self-hosting, the SIP service is deployed separately (LiveKit: telephony).
What are the licenses?
Brisvox is MIT or Apache-2.0, at your option. Smart Turn, the end-of-turn model it runs, is published under
BSD-2-Clause (repository license,
model card). LiveKit
Agents is Apache-2.0, and its turn-detector text model and the v1-mini audio model use the
separate LiveKit Model License (LiveKit: turn detector). Read both before a commercial decision.
What does LiveKit Agents document that Brisvox does not have?
From LiveKit's own documentation, as read on 14 October 2026:
- A managed cloud, LiveKit Cloud (livekit/agents). Brisvox is self-hosted only.
- Agents written in Node.js, in a separate repository, as well as in Python (livekit/agents-js).
- Its own turn-detector models, a text model and an audio model (LiveKit: turn detector).
- Agent handoff, MCP tools in Python and provider-side tools (LiveKit: tools).
- More realtime speech-to-speech models than the two Brisvox supports (LiveKit: realtime models).
- A SIP service with Twilio, Telnyx, Plivo and other trunks, and cold and warm transfer (LiveKit: telephony).
- Published sizing guidance and a default load function (LiveKit: deployments, LiveKit: server options).