The event backbone for live software

Streaks of blue light converging across a dark field

↳

Platform

The event backbone

for live software.

Streams, Replay and Observe share one ordered log, one SDK and one control plane. Everything below runs on every plan, from the free tier up.

Start building

See pricing

Streams

Ordered topics without the cluster

Create a stream with one API call. Driftwire partitions by key, keeps per-key order, and scales partitions as throughput grows — no brokers to size or rebalance.

Exactly-once delivery with idempotent producers

Per-key ordering across retries and failovers

HTTP, WebSocket and gRPC ingest on one endpoint

Schemas validated at the edge, before they spread

stream / orders

12 partitions · 41k ev/s

178,204

143,338

170,456

118,157

187,889

135,590

Replay

Rewind a consumer to the second it broke

Every consumer has its own cursor. Move it back to a timestamp, an offset or the last good deploy, and Driftwire re-delivers the exact traffic in order — at the speed you choose.

Retention up to 90 days on Scale

Replay to a sandbox consumer before going live

Throttled replays that will not swamp downstream

Every replay recorded in the audit log

consumer / billing-sync

cursor

2026-10-27 09:41:07 UTC

lag before replay

0 events

events to re-deliver

184,220

rate limit

5,000 ev/s

estimated finish

37 s

Observe

See every hop a message takes

Traces follow each event from producer to every consumer ack. Lag, error rates and slow handlers surface in one dashboard, and export to the OpenTelemetry stack you already run.

End-to-end traces with producer and consumer spans

Alerts on lag, failures and dead-letter growth

Search any event by key, ID or payload field

OTLP export to your tracing and metrics tools

trace / ord_8812

publish · api-gateway

edge ingest · fra

ordered log · append

fan-out · 3 consumers

ack · billing-sync

ack · search-indexer

Architecture

How an event moves through Driftwire

Five stages, one service. Each stage is observable and each one retries on its own, so a slow consumer never holds up the rest.

Publish

SDKs batch and sign events. Idempotency keys stop duplicates at source.

Edge ingest

The nearest region validates the schema and acknowledges in milliseconds.

Ordered log

Events append to a replicated, partitioned log with per-key order.

Fan-out

Each consumer reads at its own pace, with its own cursor and retries.

Observe

Spans, lag and failures stream to the dashboard and your OTLP collector.

Reliability

Built to be the boring part of your stack

Multi-region replication, automatic failover and a public status history. The Scale plan carries a 99.99% uptime SLA with service credits.

Delivery API · last 60 days

99.996%

60 days ago

Today

3×

Replication per region

< 30 s

Regional failover

SOC 2

Type II audited

SDKs

Idiomatic clients for six languages

Generated from one OpenAPI spec and tested against the same suite, so behaviour matches everywhere. Prefer raw HTTP? Every call is a plain REST request.

TypeScript

npm i @driftwire/sdk

Python

pip install driftwire

go get driftwire.dev/go

Rust

cargo add driftwire

Java

implementation "dev.driftwire:sdk"

Ruby

gem install driftwire

→