Bandar Systems Inc.

Developers building real-time features such as multiplayer apps, agent sessions, live broadcasts, or sandbox observability need a place to durably store and replay streaming data at the granularity of a user, device, or session. Today this typically means running and operating something like Kafka, which is heavyweight and not built to scale to unlimited per-entity streams, or bolting real-time behavior onto object storage, which is not designed for it. Either path adds operational complexity and makes it hard to let clients read historical and live events without building custom proxy infrastructure.

S2 is a serverless API that treats streams as a cloud storage primitive: you can create unlimited streams, auto-created and auto-cleaned, at any granularity such as per user or per session. Records are appended atomically to the tail of a stream and durably ordered on object storage before being acknowledged, with append latencies around 40ms or 400ms depending on tier. Reads are strongly consistent and can start from a sequence number, timestamp, or offset, with support for long-running real-time tailing via SSE or HTTP/2. The product exposes REST APIs and SDKs for TypeScript, Go, Rust, and Python, a CLI for managing basins, streams, and access tokens, and scoped access tokens so clients can connect directly and safely. An open source self-hostable version called s2-lite is also available, and the hosted system is built in Rust and stated to be SOC 2, HIPAA, and GDPR compliant.