// the find
jitsucom/jitsu
Jitsu is an open-source Segment alternative. Fully-scriptable data ingestion engine for modern data teams. Set-up a real-time data pipeline in minutes, not days
Jitsu is a self-hosted event collection pipeline — you instrument your app, events flow to Kafka or directly to your data warehouse (Postgres, BigQuery, ClickHouse, Snowflake, Redshift). It's the kind of thing you reach for when you want to stop paying Segment $500/month and have someone on the team who knows what Kafka is.
The Bulker layer is genuinely well-designed: it's a separate Go library with its own HTTP API, so you can use just the warehouse ingestion piece without pulling in the full Jitsu stack. Schema evolution is handled automatically — new fields in your event payload get added as columns without manual migrations. The test suite is thorough and uses real testcontainers (real Postgres, real ClickHouse, real Kafka) rather than mocks, which means the connector behavior you test is the behavior you get in prod. The Airbyte CDK compatibility means you can bolt on any of the hundreds of Airbyte sources without writing custom connectors.
Jitsu 2.0 is effectively a rewrite with a different architecture from Classic, but the repo still hosts both and the README doesn't clearly communicate which branch to use for what — the Classic branch is still active and the migration path is underdocumented. The functions/scripting layer (transformations in JavaScript) runs in an isolated V8 context but the security model and resource limits are not well documented, which matters if you're letting non-ops people write transforms. Kafka is a hard dependency for the production path — there's no lightweight mode that writes directly to the warehouse without a broker in between, so the ops surface area is larger than 'set up a database'. Self-hosted production deployment docs exist but are thin on observability guidance — Prometheus metrics are exposed but there's no reference dashboard.