One YAML per connector
Config, probe, mapping - one file
A single YAML defines a connector: its configuration, its capability probe, and how its catalog maps into EVE. That is why the list keeps growing.
Platform / Integration
Every source. One click.
No babysitting.
EVE connects to the databases, warehouses, object stores, and SaaS tools you already run - then keeps them in sync with log-based change capture and schema-drift policies that pause, ask, and resume. When nothing is running, it scales to nearly nothing.
29 connectors and growing · one YAML each · CDC included
01 Connect anything
Every warehouse, database, and app you run.
29 connectors today and growing every month - databases, warehouses, object stores, and SaaS. Each one is defined by a single YAML file, so new sources ship fast without a plugin release.
PostgreSQLDatabase
MySQLDatabase
MariaDBDatabase
SQL ServerDatabase
ClickHouseDatabase
SnowflakeWarehouse
BigQueryWarehouse
RedshiftWarehouse
Amazon S3Object store
Google Cloud StorageObject store
Azure BlobObject store
SalesforceSaaS
HubSpotSaaS
StripeSaaS
ShopifySaaS
Google AdsSaaS
Facebook AdsSaaS
ZendeskSaaS
+ more every monthnew sources ship fast
Auto-probing
It detects its own permissions
At connect time EVE probes what the source can actually do - read, write, CDC - and configures itself. No manual capability setup.
Sync modes
Full refresh, incremental, CDC
Pick a mode per sync, select the exact columns you want, and run it on a cron schedule. Change the shape without rebuilding the pipeline.
02 Change data capture
Real-time, and it never touches prod.
For Postgres, EVE reads the write-ahead log - the same stream your database uses for replication - so changes flow in continuously without polling or heavy queries.
- Reads the log, not the table. CDC tails the write-ahead log, so inserts, updates, and deletes stream in near real-time with no scheduled scans.
- Never mutates the source. EVE reads the replication log and writes only to your lake. Your production database is left exactly as it is.
- Preflight checks included. Before the first run, EVE verifies replication settings and permissions, so CDC starts clean instead of failing halfway.
- Full change fidelity. Every change arrives in order and lands as an append-only history in open Parquet.
pg_orders · log-based CDC
How a change travels
source dbreplication logEVE lake
Source
orders table, unchanged
read-only - EVE never writes back
Replication log
WAL - ordered change stream
inserts, updates, deletes in order
EVE lake
Parquet in your own S3
near real-time, preflight-checked
03 Schema drift
Schema drift won't break prod.
Every run starts with a discover-time diff. When a source changes shape, your policy decides what happens next - and a breaking change never slips through silently.
DETECT
Diff before every run
At discover time, EVE diffs the incoming source schema against the last known one - before a single row moves.
discover > diff > decideDECIDE
A policy per sync
Choose auto-propagate, manual approve, or stop. A breaking change pauses the pipeline and freezes the catalog until a human approves.
policy: manual-approveRESUME
One click, back in flow
Review the diff, accept it, and the pipeline resumes exactly where it paused. Nothing silently breaks downstream.
accept + resume04 Cost
It costs nothing when idle.
Ingestion is not a fleet you keep warm. When no sync is running, the infrastructure suspends itself and resumes on demand.
~$0
Idle infra suspends itself
When no sync is running, ingestion infrastructure scales all the way down. You are not billed for compute that is waiting around.
On demand
Wakes on the next run
The next scheduled or manual sync brings ingestion back automatically. No warm-up toil, no standing cluster to pay for.
Storage only
Suspended means storage cost
A suspended source costs only its Parquet in your S3 bucket. Open format, your bucket, no export tax to leave.
Behavior measured on our own staging environment, not synthetic benchmarks.
Get started
See your data on EVE.
A 30-minute demo on your use case, or a free workspace on your own isolated cluster.