Open Source Reactive SQL Engine

Extract. Transform. Load.
One binary. Zero infrastructure.

Connect any API or database. Transform with composable SQL pipelines. Deliver results anywhere. Replace Flink, Fivetran, and ksqlDB with a single binary.

Download LiteJoin View on GitHub

brew install litejoin/tap/litejoin

Extract

Stripe CDC 42/s

postgres_cdc

Weather API 12/s

http_api

Users DB 38/s

postgres_cdc

Transform

order_enrichment 38/s

Join

chain

revenue_5m 12/eval

Window 5m

Load

SSE Out 38/s

sse

Postgres 12/s

postgres

Webhook 38/s

http

How it works

Three verbs. That's the whole product.

Connect sources, transform with SQL, deliver to sinks. The mental model is simple because the engine handles the hard parts.

01 — Extract

Connect your data sources

Paste a URL or connection string. LiteJoin polls APIs with intelligent diffing, streams CDC from Postgres, and consumes from Kafka.

REST API poll + diff

Postgres CDC logical replication

Kafka consumer groups

HTTP Webhook POST ingest

02 — Transform

SQL on live streams

Join streams, aggregate with time windows, or write arbitrary SQL. Transformations are composable — chain them together, fan them out.

⇌ Joins multi-stream

⏱ Windows tumbling / session

* Custom SQL full SELECT

↩ Chaining composable

03 — Load

Deliver results anywhere

Push transformed data to Postgres, fire webhooks, stream via SSE to your frontend. Delivery health and retry built in.

HTTP Webhook POST

SSE real-time push

Postgres upsert

SQLite embedded

Kafka async producer

Composable Transforms

One source. Many pipelines.

LiteJoin transformations fan out and chain together. One Stripe CDC source can feed a join, a window, and a SQL filter simultaneously. No linear pipeline limitations.

→

Fan-out

One source feeds multiple transformations independently

↩

Chaining

Output of one transformation becomes input to the next

⨆

Multi-input

Multiple sources feed into a single join transformation

Transformation — order_enrichment (Join)

SELECT c.key AS charge_id, json_extract(c.payload, '$.amount') / 100.0 AS amount, json_extract(c.payload, '$.currency') AS currency, json_extract(u.payload, '$.name') AS customer_name, json_extract(u.payload, '$.plan') AS plan_tier FROM stripe_charges c LEFT JOIN users u ON json_extract(c.payload, '$.customer') = u.key -- Chains into: revenue_5m (Window 5m tumbling) -- Fan-out to: high_value_filter (SQL, amount > 1000)

MCP Server

Build pipelines from your editor

LiteJoin ships an MCP server that exposes every engine capability as a tool. Create sources, chain transformations, wire sinks — all through natural language in your IDE.

Describe an entire ETL pipeline in one sentence. The MCP creates sources, writes the SQL, chains transformations, and starts the pipeline. Open Studio to watch it run.

Works with any MCP client:

Claude Code Cursor Windsurf Claude Desktop

Claude Code

> Join Stripe charges with users, compute 5-min revenue by plan, alert via webhook when any plan drops 20%. Creating source stripe_charges (postgres_cdc) Creating source users (postgres_cdc) Creating join order_enrichment ON $.customer = users.key Chaining window revenue_5m (tumbling 5m) GROUP BY plan_tier Chaining SQL drop_detector compare current vs previous window Creating sink webhook_alert (http) Pipeline running 2 sources, 3 transforms, 1 sink Open Studio to visualize the pipeline

LiteJoin Studio

See your data flow

The visual companion to your pipelines. Watch data stream through every transformation. Ships with the binary — no account, no signup.

Dataflow Canvas

Your entire pipeline as a left-to-right DAG. Sources, transformations, sinks. Edges animate with live throughput.

Live Data Inspection

Click any node to see live records streaming. Double-click a row to inspect the full JSON payload.

SQL at Every Step

Every transformation exposes its SQL. Edit inline, run it, see results immediately. Write it or let MCP write it for you.

Ambient Observability

Throughput, latency, error counts on every node. Edge thickness shows volume. No separate dashboard needed.

Fan-out & Chaining

The canvas makes composition visual. See fan-out relationships, inspect data at every stage of a multi-step pipeline.

MCP Companion

Build pipelines from Claude Code. Open Studio and the canvas updates live — the dashcam to your MCP workflow.

Ships with the binary. No account needed. Just run litejoin studio

Why LiteJoin

You shouldn't need a platform team
for real-time data

Purpose-built for developers who want real-time ETL without enterprise operational weight.

	LiteJoin	Kafka Streams	Apache Flink	Fivetran
Time to first pipeline	5 minutes	Hours	Days	Minutes
Infrastructure	Single binary	Kafka cluster	Flink + Kafka	SaaS
API to stream	Paste a URL	Kafka Connect	Custom source	Connectors
Transformations	SQL (composable)	Java DSL	Java / SQL	UI / dbt
Composability	Fan-out + chaining	Linear topologies	DAG (complex)	Linear
AI-native (MCP)	Built in	No	No	No
Cost	$0 open source	$$$$	$$$$	$$$$
Best for	App developers	Java backend teams	Data engineering	Non-technical

Connectors

Connect to anything. Stream everywhere.

Sources and sinks — with more shipping continuously.

Sources

REST API

Postgres CDC

Kafka

HTTP

MySQL

MongoDB

Sinks

Webhook

SSE

Postgres

SQLite