Snakream

Application state often ends up split across process memory, custom recovery code, and one-off real-time plumbing. When a worker crashes or a client reconnects, replay and recovery get harder than they should be. Snakream provides durable, replayable streams for application state. Give each document, session, task, or run its own stream for recovery, replay, and live updates over plain HTTP.

What you get

Crash recovery. Replay an object’s stream and resume from a known state after a restart. Live updates over HTTP. Read from the beginning, catch up, or tail live with long-poll or SSE. curl and EventSource work out of the box. Automatic lifecycle. Keep recent data hot, archive older data automatically, and expire finished streams with TTL when they no longer matter.

Try it now

# Create a bucket and stream
curl -X PUT https://stream.tonbo.dev/ds/demo
curl -X PUT https://stream.tonbo.dev/ds/demo/hello

# Append data
curl -X POST https://stream.tonbo.dev/ds/demo/hello \
  -H 'Content-Type: application/octet-stream' \
  --data-binary 'hello world'

# Read from the beginning
curl 'https://stream.tonbo.dev/ds/demo/hello?offset=-1'

# Subscribe for live updates (SSE)
curl 'https://stream.tonbo.dev/ds/demo/hello?offset=-1&live=sse'

Learn more

Quick Start — step-by-step walkthrough
Why Snakream — where it fits, and where it doesn’t
Concepts — streams, buckets, offsets, read modes, snapshots
Architecture — how the cluster works
API Reference — full HTTP API surface
Protocol Spec — the Durable Streams Protocol and extensions

Getting Started

Concepts

Architecture

API

Protocol Specification

Comparisons

What you get

Try it now

Learn more

​What you get

​Try it now

​Learn more

What you get

Try it now

Learn more