Why not Debezium?
Debezium is a great change-capture engine, but it's built to sit in front of a streaming platform — Kafka and Kafka Connect — and to run as separate Debezium Server instances you have to configure and babysit. Fabric runs the CDC readers in-process as native connectors. There's no streaming platform to provision, no Maven jars on the restart path, and no application.properties to manage — configuration lives in the database.
Why not Kafka?
Kafka gives you a durable message bus, but for "database changes into a warehouse" it adds brokers, a schema registry, Kafka Connect, and a team to keep them healthy. Fabric removes all of that: each source is one supervised process writing straight to your destination. If you already run Kafka, Fabric is a simpler path to the same result.
Why not Fivetran?
Fivetran is a hosted service — your data goes through their cloud, and you pay per row. Fabric comes both ways: self-hosted, it runs in your own VPC against your own object storage, so your data stays in your control; on the hosted plans your rows land in your own destination account and the runners hold them only until the commit. The footprint is tiny, and there's no per-row vendor lock-in.
Why only these sources?
Sources and destinations are registered behind a common connector SDK, so adding a new database or a new warehouse is implementing one interface — no rewiring of the pipeline. The set today is what's wired end-to-end; more are on the roadmap.
What's the license?
Fabric is licensed under the Business Source License 1.1. You may use, modify, and redistribute it — including in production for your own use — but you may not offer it to third parties as a managed / hosted service. After the change date it re-licenses to the Apache License 2.0.