Live production alert  ·  Pub/Sub lag 34s
The Pub/Sub connector is experiencing elevated lag. Zipline is retrying at 2× throughput. No messages at risk - exactly-once guarantees active.
v2.4  ·  Exactly-once for Kafka 3.8

Stream every row,
to anywhere, in
under a second

Zipline connects your production databases to Kafka, PubSub, Snowflake, BigQuery, and S3 with exactly-once delivery and sub-second latency - no data loss, no duplicates. Apply the spec, then bind it to a runtime - no dashboards, no wizard.

SOC 2  ·  Apache 2.0  ·  self-hosted
View A · section A-AScale 1:1
142 ms
Commit → sink ack · P50
ONE ROW ±0 rows EXACTLY ONCE WAL ORDER PRESERVED PER KEY WAL-SOURCED · NO DUPLICATES ACROSS RESTARTS
View B · detail2:1
routes:
  - from: orders-db
    to: [kafka-out]
orders.yaml · 22 lines
View C · revisionsCurrent 2.4
  1. 2.4  exactly-once, Kafka 3.8
  2.        2.3  schema evolution
  3.        2.2  replay windows
  4.        2.1  Terraform provider
View D · bill of materialsShipping
ItemPartDirClass
01postgressrc · snkCDC
02mssqlsrcCDC
03kafkasnkBus
04filesnkFile
27 further parts continued on sheet 2
View E · approvalAudited
SOC 2 TYPE II HIPAA BAA · GDPR ALIGNED FEDRAMP IN PROCESS
View F · performanceSingle node · m5.4xlarge
ZIPLINE DBZ + CONNECT 2.1M evt/s ~180K 100K 500K 1M 2M EVENTS / SEC · LOG SCALE · EQUIVALENT DURABILITY
View G · deployScale · seconds
APPLY → FIRST ROW 0 60 30 s NO CLUSTER · ONE BINARY
Zipline · delivery guarantee Rev 2.4 Sheet 1 of 2
Trusted by engineering teams at
Stripe Shopify Figma Notion Vercel Cloudflare
"Figma's design data model is complex - hundreds of table types, high write volume, very low tolerance for inconsistency. Zipline handles the cardinality and throughput without breaking a sweat. The Terraform provider made our rollout completely automated."
Principal Engineer · Figma · Backend Infrastructure
How it works

Three steps.
One file.

orders.yaml one file · three steps
name: orders
connectors:
  - name: orders-db
    type: source/postgres
    host: "${POSTGRES_HOST}"
    port: "${POSTGRES_PORT}"
    database: "${POSTGRES_DATABASE}"
    user: "${POSTGRES_USER}"
    publication: "${POSTGRES_PUBLICATION}"
    slot_name: "${POSTGRES_SLOT_NAME}"
    secret_refs:
      - { field: password, ref: postgres-password }
  - name: kafka-out
    type: sink/kafka
    brokers: "${KAFKA_BROKERS}"
    topic_prefix: "${KAFKA_TOPIC_PREFIX}"
routes:
  - from: orders-db
    to: [kafka-out]
~ zipline pipeline apply -f orders.yaml✓ applied orders  the spec is recorded. nothing streams yet.~ zipline pipeline bind orders --instance src-7f3abinding src-7f3a to orders...✓ bound orders <- src-7f3a  state: pending  route orders-db → kafka-out · running
Lag P50
142ms
Rows
1.2M
Lost
0
Performance

Numbers that matter
in production.

2.1M
events / sec · single node
Sustained throughput on a single m5.4xlarge. No cluster required to start.
2M
<1s
end-to-end latency · P50
Median latency from DB commit to sink acknowledgement, measured continuously in prod.
ms
12×
vs Debezium + MSK
Benchmark vs Debezium on Amazon MSK at equivalent durability settings and message sizes.
12
Use cases

What teams point it at.

See all
What Zipline does

Everything you need.
Nothing you don't.

Exactly-once guaranteed
Transactional WAL-based CDC with idempotent sinks. No duplicates, no missed events - even across network partitions and process restarts.
Schema evolution
Column additions, renames, and type changes propagate automatically to downstream sinks - with configurable compatibility modes and alerting.
Time travel & replay
Replay any event window from WAL history. Backfill a new sink, debug a downstream bug, or rebuild a derived dataset from scratch.
Enterprise-grade access
SSO, RBAC, audit logs, and field-level encryption. SOC 2 Type II certified. HIPAA Business Associate Agreements available for healthcare teams. Run it in our cloud or entirely in your own infrastructure.
Try the CLI

A real simulator.
Nothing touches a runtime.

zipline playground
zipline playground - a simulated CLI. nothing here touches a real runtime.
you are already enrolled against a demo runtime. try `zipline status`.
`help` lists what is simulated, `tour` lists the guided tours. refresh to reset.
$
Connectors

30 sources & sinks,
ready on day one.

View all connectors
Source · RDBMS
Postgres
Source · RDBMS
MySQL
Source · Document
MongoDB
Sink · Streaming
Kafka
Sink · Streaming
PubSub
Sink · Warehouse
Snowflake
Sink · Warehouse
BigQuery
Sink · Object Store
S3
Source · Cache
Redis
Sink · Search
Elasticsearch
Sink · Streaming
Kinesis
30 connectors total
View all →
Pricing

Start free. Scale when ready.

Community
$0
forever · self-hosted
  • Unlimited pipelines
  • Up to 500K events/day
  • All connectors included
  • Community Slack support
  • Apache 2.0 license
Get started free
Enterprise
Custom
volume pricing · dedicated support
  • Everything in Team
  • SOC 2 Type II + HIPAA
  • FedRAMP path available
  • Dedicated Slack channel
  • 99.99% SLA + 1h response
Talk to sales

Your data isn't waiting.
Neither should you.

Connect your first database in under 30 seconds.
No credit card. No lock-in. Apache 2.0 licensed.