Problem

S2.dev addresses the infrastructure gap between durable event streaming and fine-grained, real-time application state. Traditional systems such as Kafka and Kinesis generally rely on a small number of provisioned, durable topics, while systems such as Redis Streams and NATS offer more granular state at the expense of durability. This creates operational complexity and poor economics when applications need many independent streams, particularly for AI-agent sessions, user journeys, or parallel workflows. S2.dev cites Kinesis at $29 per month per stream versus roughly $0.05 per GiB with S2, making the cost difference especially relevant for high-cardinality workloads.

The killer use case is long-running AI-agent execution. Developers need to stream tokens, progress, tool calls, and events to a frontend in real time while preserving the complete session for replay, auditing, recovery, and coordination among agents. With conventional infrastructure, dropped connections can lose data or break streams; S2.dev's target is to make each agent session a durable, independently addressable stream so network hiccups and multiple readers do not compromise the workflow.

Product / Service

S2 is a serverless API for unlimited, durable, ordered record streams. Applications append records, tail streams in real time, and replay them from any retained point, using the service as a cloud-storage-like primitive for streaming data rather than managing brokers, partitions, connection infrastructure, or custom WebSocket and SSE systems. The platform supports agent sessions, resumable LLM token streaming, live job views, event synchronization, broadcast data feeds, and observability streams.

The delivery model is usage-based and managed by S2.dev, with transparent per-GiB transfer pricing, a stated 99.99% availability SLO, and no need to provision streams individually. The intended benefit is to combine the simplicity and scalability associated with object storage with real-time delivery: a producer can publish once, while many consumers independently follow live updates or replay history at their own pace.

Market

S2.dev competes in cloud data infrastructure, specifically serverless event streaming, durable real-time streams, and infrastructure for AI-agent state and application synchronization. Its closest architectural alternatives are Kafka and Kinesis for durable streaming, NATS and Redis Streams for lower-latency ephemeral state, and bespoke WebSocket or SSE backends for live application experiences. S2.dev differentiates by offering a potentially unlimited number of granular streams with durable retention, direct internet access, and replayability rather than forcing developers to choose between a few durable firehoses and many less-durable state channels.

The company is no longer merely pre-launch: S2.dev announced general availability on February 25, 2026, after running production workloads for customers, and raised a $3.85 million seed round led by Accel with participation from Y Combinator, bringing total capital raised to $5.5 million. Public customer evidence includes production testimonials from Kernel, which uses S2 for real-time telemetry, and Trigger.dev, which uses it for long-running AI sessions. However, public revenue, customer-count, and usage figures were not disclosed, so the available evidence supports early production traction rather than a quantified scale claim.