Mooncake Is Joining Databricks (www.mooncake.dev)

Mooncake announced it’s joining Databricks, bringing its pg_mooncake technology to unify Postgres-based OLTP apps with Lakehouse analytics and AI. The move aims to eliminate the usual tradeoffs of HTAP by letting developers keep Postgres as the transactional source of truth while syncing that data directly into a Lakehouse for analytics and model workloads—no custom ETL, Debezium/Kafka/Flink CDC pipelines, or duplicated engineering effort. Mooncake’s team frames this as making the “Lakebase” (Postgres) and Lakehouse feel like one platform, now under Databricks’ umbrella. Technically, pg_mooncake remains open source and provides a Postgres extension that synchronizes data into Lakehouse engines (Spark, Snowflake, DuckDB, Trino and Databricks’ own agent bricks), enabling apps to serve intelligence from Postgres while analytics and agents operate on the same live data. That design promises simpler operational stacks for startups and enterprises, faster ML/analytic feedback loops, and support for agent-scale workloads (thousands of concurrent agents) without manual pipelines. For the AI/ML community this reduces data plumbing friction, speeds product-to-model integration, and centralizes tooling for both transactional and analytical/agent-driven use cases.