> Device-to-device connectivity for edge AI
Encrypted NAT-traversing streams between cameras, robots, and GPU workers. Open source.
let session = pb.connect("camera-1").await?;
let mut video = session.open_stream("4k-feed").await?;
let mut events = session.open_stream("inference").await?;
// Stream initialized. Direct link established.
loop {
let frame = video.next().await?;
process_frame(frame);
}Punch through symmetric NATs using STUN/TURN. Direct IP-to-IP delivery with sub-50ms latency. Relay fallback when direct fails.
Open multiple named streams per session. Video, telemetry, and control on one encrypted connection. No port management.
Run your own control plane. No vendor lock-in. One binary, one config file. Or use managed cloud — your call.
01. HOW IT WORKS
Identity
Each device gets a persistent cryptographic identity and registers with the control plane by name.
Discovery
Find peers via the API and connect using name-based routing. NAT traversal is automatic.
Streaming
Open named bidirectional streams and pipe any data. Video, sensor, control — all multiplexed.
02. USE CASES
Built for demanding high-bandwidth environments where latency and security are non-negotiable.
03. PRICING
Peerbase is free to use during the beta period. Managed cloud and self-hosted — both fully available, no limits.