What you need
2 projectsCommercial useMaintainedSetupCost
- Commercial use
- Yes
- LGPL-3.0. You can host it for paid classes without restriction. If you modify its own code and distribute it, those changes must be open.
- Maintained
- Active
- Last push on 2026-10-08. The repository is not archived.
- Setup
- Hard
- It needs a clean, dedicated Ubuntu 22.04 server that does not share ports 80 and 443 with another site, a domain with a valid SSL certificate (browsers will not grant camera and microphone access without one), and UDP ports 16384 to 32768 open. Installation is through the official bbb-install.sh script.
- Cost
- High
- The software is free; the server is the cost. The stated production minimum is 8 cores, 16 GB RAM and 250 Mbps symmetrical bandwidth, with 500 GB of disk if you record sessions or 50 GB if you do not.
- What to watch out for
- What 250 Mbps of symmetrical bandwidth costs on a Chinese cloud has not been checked here; get a quote before deciding. The docs also ask for both an IPv4 and an IPv6 address.
- Commercial use
- Yes
- LiveKit Server, Egress and the Agents framework for voice AI are all Apache-2.0: commercial use and modification are allowed, and you do not have to publish your changes.
- Maintained
- Active
- Last push on 2026-10-15. The repository is not archived.
- Setup
- Moderate
- LiveKit Server: an official config generator (the livekit/generate image) produces Caddy, Redis, docker compose and a startup script in one go; run the script on a Linux VM and it is up. You need your own domain: the main domain and the TURN domain must both point at the machine, and self-signed certificates do not work (Caddy obtains them automatically). Open ports 80, 443, 7881 (TCP), 3478 (UDP) and the whole UDP range 50000 to 60000. Egress: the generator does not cover it. Run it as a separate Docker service (livekit/egress) with the server’s URL and API key, connected to the same Redis as the server. Full-room recording runs Chrome inside the container, which needs the SYS_ADMIN capability or Chrome fails to start. Recordings go to S3, Azure or Google Cloud Storage. Both have official Helm charts for Kubernetes.
- Cost
- Varies
- Self-hosting is free; the money goes on two kinds of machines. LiveKit Server is bound by CPU and bandwidth: in the official benchmark, one 16-core Google Cloud c2-standard-16 handled a 150-person video meeting where everyone sees everyone, at 85% CPU, and the docs recommend compute-optimized machines and 10 Gbps networking for production. Each Egress instance should have at least 4 CPUs and 4 GB; a full-room recording takes 2 to 6 CPUs, and one instance records one room at a time, so you need as many instances as rooms you record at once. If you would rather not run it all, LiveKit Cloud has a free tier with 5,000 participant minutes and 50 GB of transfer a month; paid plans start at $50 a month, with overage at $0.0005 per participant minute and $0.12 per GB.
- What else you need
- LiveKit Server is required: it runs the rooms and routes the media. Redis is recommended for production and required once you add Egress; both must use the same one. Egress records meetings to MP4 or HLS, or live-streams them over RTMP to platforms like YouTube Live and Twitch; recordings need an S3, Azure or Google Cloud bucket. You also need a domain of your own. Ingress (bringing outside RTMP streams into a room) and Agents (voice AI) are optional and not needed for meetings alone.
- What to watch out for
- The meeting interface, login and permissions are for your own developers to build. The official sample app, LiveKit Meet, is open source, but its README only covers local development, not production. The Chrome inside Egress runs with its sandbox off by default; turning it on takes an extra setting and a seccomp profile.
Your industry is not here?
Tell me the industry and what you need solved. I will look at it next.