4.2 billion events moved yesterday

4.2B events moved yesterday

Every event, from every source, delivered exactly once.

Real time pipelines from 400+ sources to your warehouse, lake or app. No duplicates, no gaps, no cron jobs.

moving data for

  • Northwind

  • HALCYON

  • kestrel

  • Parabola

  • Oxbow&Co

  • VANTAGE

  • meridian

  • Northwind

  • HALCYON

  • kestrel

  • Parabola

  • Oxbow&Co

  • VANTAGE

  • meridian

  • Northwind

  • HALCYON

  • kestrel

  • Parabola

  • Oxbow&Co

  • VANTAGE

  • meridian

How it works

Three stages. Zero babysitting.

Watch an event travel from a Postgres row to a warehouse table. Every hop is ordered, retried and accounted for.

01Capture
02Transform
03Deliver
postgres.orders
1,204 ev/s
stripe.charges
312 ev/s
kafka.clicks
9,870 ev/s
webhooks.app
88 ev/s
sdk.web
2,410 ev/s
mongo.users
64 ev/s
snowflake.analytics
ordered · exactly once
bigquery.events
ordered · exactly once
s3://lake/raw
ordered · exactly once
clickhouse.metrics
ordered · exactly once
webhook.billing
ordered · exactly once
mask(email)ada@acme.com → a••@acme.com
where amount > 02 events filtered, logged
dedupe on id3 retries → 1 write
6 sources13,948 events/s in
0.4 msadded per event
180 ms p990 duplicates, 0 gaps

Exactly once

Sent three times. Landed once.

Apps retry. Networks hiccup. Sluice makes sure every event still lands exactly one time.

Your app
Sluice
Warehouse
1
Your app retries
A timeout makes your app send the same order three times. Totally normal.
2
Sluice spots the repeats
Every event carries an ID. Sluice has seen this one, so the copies are skipped.
3
One clean row lands
Your warehouse gets the order exactly once. No dedupe query needed.
order #1042
try 1
order #1042
try 2
order #1042
try 3
Go on, retry it. It still lands once.
0 sent · 0 skipped · 0 written

Platform

Built for pipelines you can't afford to break.

The boring parts of streaming, done properly. So the data team can go back to answering questions.

Replay any window

Rewind a pipeline to any point in the last 30 days and backfill without touching production.

Replaying from Nov 18, 09:36
Nov 01Nov 08Nov 15Nov 22Nov 30
◀◀30d
backfill 2.1m events

Schema drift, handled

New columns flow straight through. Breaking changes pause one table, not the pipeline.

schema · ordersv41→···
idstring
amountdecimal(12,2)
currencychar(3)
+discount_codestringnullable
created_attimestamp

Lag you can see

Throughput, lag and errors per pipeline, with a trace for every row.

p50142 ms
p99180 ms
−60 min−45−30−15now

Alerts that say what to do

Sluice pauses the broken table, tells you why in Slack, and resumes with one click.

# data-alerts3 members
SluiceAPP10:42 AM
Paused 1 table in orders_v2. Breaking change detected.
- amount int
+ amount string
0 events lost · 41,203 buffered
Review diff
Resume
SluiceAPP10:51 AM
Resumed. Backfill finished: 41,203 events, 0 duplicates.

live · events delivered exactly once today

p99 latency

180 ms

uptime

99.99%

connectors

400+

duplicates

0

Developer first

Ship a pipeline before your coffee cools.

Pipelines are twelve lines of YAML. Review them in pull requests, deploy from CI, preview every branch.

TypeScript

Python

Go

Terraform

sluice.yaml
pipeline: orders_to_warehouse
source:
type: postgres
tables: [orders, refunds]
transform:
sql: select * except (card_last4)
destination:
type: snowflake
delivery:
guarantee: exactly_once
zsh · sluice deploy
$
● streaming 12,408 events/s lag 142 ms
$

Connectors

400+ connectors. One guarantee.

Every connector ships with exactly once delivery, schema handling and replay. Missing one? Build it in an afternoon with the SDK.

PG
source + dest
Postgres
CDC · logical replicationConfigure →
MY
source
MySQL
CDC · binlogConfigure →
MG
source
MongoDB
change streamsConfigure →
KF
source + dest
Kafka
topics · consumer groupsConfigure →
ST
source
Stripe
webhooks + backfillConfigure →
SF
source
Salesforce
bulk + streaming APIConfigure →
HS
source
HubSpot
objects · eventsConfigure →
SH
source
Shopify
orders · inventoryConfigure →
SN
destination
Snowflake
snowpipe streamingConfigure →
BQ
destination
BigQuery
storage write APIConfigure →
DB
destination
Databricks
delta · unity catalogConfigure →
CH
destination
ClickHouse
native protocolConfigure →
S3
destination
Amazon S3
parquet · icebergConfigure →
IC
destination
Iceberg
REST catalogConfigure →
PC
destination
Pinecone
embeddings · upsertsConfigure →
+
sdk
Build your own
sluice connector newConfigure →

Customer story

“We deleted 40 dedupe jobs the week we moved to Sluice. Finance finally trusts the revenue dashboard, and nobody gets paged at 3am for a stuck pipeline anymore.”

“We deleted 40 dedupe jobs the week we moved to Sluice. Finance finally trusts the revenue dashboard, and nobody gets paged at 3am for a stuck pipeline anymore.”

“We deleted 40 dedupe jobs the week we moved to Sluice. Finance finally trusts the revenue dashboard, and nobody gets paged at 3am for a stuck pipeline anymore.”

Maya Lindqvist

Head of Data, Kestrel

40

40

40

dedupe jobs removed

6x

6x

6x

faster dashboards

$310k

$310k

$310k

saved per year

Pricing

Pay for events, not seats.

Start free and scale with usage. Every plan includes exactly once delivery.

Monthly events
50M
Team plan
$249/ month
includes 50M · $4 per extra 1MStart 14 day trial
◀▶
Freeup to 1M
Team1M to 50M
Enterprise50M and beyond

Free

$0

$0

$0

/ month

For side projects and your first pipeline.

1M events a month

5 connectors

7 day replay

Community Slack

Team

most teams pick this

$249

$249

$249

/ month

For teams running production data on Sluice.

50M events, then $4 per 1M

All 412 connectors

Exactly once delivery

30 day replay

Support in your Slack

Enterprise

Custom

Custom

Custom

For regulated, high volume workloads.

Unlimited events

Run in your own cloud

SSO, SCIM, audit logs

99.99% uptime SLA

A named engineer

FAQ

Questions, answered.

AK

ML

JR

Still stuck? Ask an engineer.

median reply 11 min

Every event lands in your destination one time, even through retries, network splits and redeploys. Sluice keys each event on its source and ID, and commits the write and the offset together.

idempotency_key = sha1(source, event.id)Read how dedupe works →

Free up to 1M events a month

Stop moving data twice.

Create a free website with Framer, the website builder loved by startups, designers and agencies.