Data mapping
Tracks which source tables map to which target tables, column by column, with explicitly defined type coercions.
MAX LATENCY
15.4ms
↓ 12% vs last hour
Real-time data replication · v1.0
Moving change events into modern targets should be fast, boring, and lock-in free.
JetStreams is an API-first, Python-backed real-time data replication engine built for the modern data stack. It replaces legacy data replication systems, stripping away heavy license fees and audit complexity to turn existing databases into verified pipelines in minutes.
JETSTREAMS / live · stream-01 · region us-east-1
CDC · row-level diff · sub-second
Platform · API-first
Everything is an API call.
No code edits. Re-aiming the product at a different client deployment is a configuration change, not a redeploy. The same surface area drives the web console, your CI/CD jobs, GitOps controllers, and any automation you already operate.
~/jetstreams · provision a pair
$ curl -X POST https://api.jetstreams.net/v1/pairs \
-H "Authorization: Bearer $JS_TOKEN" \
-d '{
"source": { "kind": "oracle", "host": "ora-prod", "schema": "HR" },
"target": { "kind": "snowflake", "warehouse": "ANALYTICS_WH" },
"mode": "cdc",
"scope": { "tables": ["EMPLOYEES", "DEPARTMENTS"] }
}'
→ scope-analyzer examining 12 tables, 38 columns, 6 FKs
→ datatype-map NUMBER(38) ▸ NUMBER, VARCHAR2 ▸ VARCHAR ............ ok
→ source-grants least-privilege user provisioned ...................... ok
→ connector registered with JetStreams-connect ...................... ok
→ verify 5,000 rows backfilled, 0 row-level diffs ............... ok
pair_id: ora-hr.snowflake-analytics ● LIVE
Reference architecture
Web console
CI / CD
GitOps
Automation
▼
JetStreams API
REST · OpenAPI
Python replication engine
JetStreams-connect
Five-step execution
From raw credentials to a verified, streaming pipeline.
The five-step engine automates the heavy lifting of a real CDC rollout. You hand it source and target credentials. It returns a pipeline that has already proven itself against a row-level diff before any human is asked to trust it.
01
Create
Generates the replication user with the exact least-privilege grants for the source.
02
Configure
Automates configuration of source and target schemas, including datatype coercions.
03
Register
Registers the streaming components on both ends and confirms healthchecks before moving on.
04
Verify
Performs a precise, row-level diff to verify the end-to-end flow against real production rows.
05
Expose
Lights up the live operations dashboard so on-call can watch lag and throughput from one place.
Pre-flight precision · Scope Analyzer
Diagnostic intelligence before a single byte moves.
Migrations fail when pipelines break mid-flight over schema quirks. The Scope Analyzer examines schemas, foreign keys, and datatypes on both ends and surfaces the problems that quietly stall most proofs of concept. The analyzer is plugin-based, so a new source-to-target pairing is a new analyzer, not a rewrite.
Scope analyzer · plugin model
Issue discovery
Identifies critical friction points like NOT NULL violations, length overflows, charset mismatches, and numeric precision loss.
Datatype mismatches
Flags necessary pre-cast work like Oracle NUMBER(38) to Snowflake, VARCHAR2 primary keys, TIMESTAMP precision, and large objects.
Unsupported datatypes
Surfaces un-streamable types up front like XMLType and JSON Duality, so teams know early what requires manual intervention.
Verified source-to-target matrix
Eight verified, end-to-end pairings.
Every pairing on this list has passed the same five-step pipeline against real production-shape data. New pairings come online through the platform's scope-analyzer plugin model rather than as bespoke connectors.
| Source | Target | Status | |
|---|---|---|---|
| O Oracle | › | SF Snowflake | Available |
| O Oracle | › | P PostgreSQL | Available |
| O Oracle | › | M MySQL | Available |
| O Oracle | › | DB Databricks | Coming soon |
| P PostgreSQL | › | O Oracle | Available |
| P PostgreSQL | › | P PostgreSQL | Available |
| M MySQL | › | O Oracle | Available |
| M MySQL | › | M MySQL | Available |
| SQ SQL Server | › | P PostgreSQL | Available |
Adaptive deployment patterns
You own the infrastructure. JetStreams adapts to your network shape.
Same product, same API surface, three shapes. Pick the topology that matches your security model and the realities of where your databases already live.
PATTERN 01
Source-side agent
JetStreams runs inside the same VPC as the source database. The target is reached over your existing outbound path.
PATTERN 02
Target-side agent
Ideal for client-owned cloud environments. JetStreams sits next to the target and pulls from the source over an authenticated path.
PATTERN 03
Hub deployment
The simplest footprint for managed services. One JetStreams hub fans in from many sources and out to many targets.
Operations dashboard
Total visibility at a one-second cadence.
Sustained latency, per-target lag, throughput, and deep technical readouts in one operator-facing view. The same dashboard that ships with the platform is what your on-call sees, with no separate observability stack to wire up.
REPLICATION LAG
12ms
p99 across 8 pairs
THROUGHPUT
3.0k rows/s
sustained 60s window
ERROR RATE
0.01%
● within SLO
Architecture diagram · live
Technical readout · per pair
Pair
Latency
Throughput
ora-prod ▸ snowflake
14 ms
2.3 MB/s
ora-prod ▸ postgres
9 ms
0.8 MB/s
pg-hr ▸ postgres-dw
7 ms
0.4 MB/s
mysql-shop ▸ mysql-ods
11 ms
1.6 MB/s
mysql-shop ▸ oracle
23 ms
0.6 MB/s
Controlled security blast radius
Single-tenant. Self-hosted.
Every deployment is single-tenant and 100% self-hosted. Nothing phones home to a vendor service. Analogous least-privilege modes exist for PostgreSQL and MySQL, so the source database hands JetStreams exactly the rights it needs and no more.
- 01Pluggable database mode. Tighter security blast radius, easier compliance review, fewer grants.
- 02Multi-PDB mode. Cross-pluggable database capture via a single common user.
- 03Encrypted registry. Configuration metadata is AES-GCM encrypted at rest with a customer-held master key.
PDB-local mode
PDB
SCOPED CAPTURE
Multi-PDB mode
CONTAINER · CDB
PDB 1
PDB 2
PDB 3
COMMON-USER CAPTURE
Platform roadmap
Expanding the replication horizon without compromising the API-first architecture.
Now · v1
Eight verified pairings
Oracle, PostgreSQL, MySQL, and SQL Server as sources. Snowflake, PostgreSQL, MySQL, and Oracle as targets.
Late-stage testing
Oracle to Databricks
Currently in late-stage internal testing against real production-shape data on the scope-analyzer plugin model.
Q3
Same-source pairings
PostgreSQL and MySQL to Snowflake. Same engine, same dashboard, more rooms to land your data in.
Q4
Enterprise features
Lineage, alerting, schema-evolution policies, audit export, and an opt-in managed cloud offering for teams that prefer not to self-host.
Available for pilots
Turn existing databases into verified pipelines in minutes.
Pilots run on your infrastructure against your real databases. We hand you back a working pipeline, a runbook, and an operations dashboard.