Show HN: SerenDB – A Neon PostgreSQL fork optimized for AI agent workloads (github.com)

SerenDB is an open-source, serverless fork of PostgreSQL aimed at replacing the traditional monolithic storage layer with a distributed storage engine and stateless Postgres compute nodes — positioning itself as an alternative to AWS Aurora Postgres and marketed for AI agent workloads. It separates storage and compute: pageservers provide scalable cluster storage while “safekeepers” offer a redundant WAL service that durably records write-ahead logs until pageservers process them and upload to cloud storage. Compute nodes are standard PostgreSQL instances (patched) that connect to that storage, so you can use psql or any Postgres client and run many lightweight, isolated endpoints. Technically, SerenDB supports copy-on-write timelines and cheap branching (create independent Postgres endpoints on branches), making it attractive for AI/ML use cases that need many ephemeral databases for model training, A/B migration tests, or parallel agent simulations. The project is implemented with Rust tooling (requires protoc 3.15+, Rust toolchain pinned), provides local dev scripts and a free tier to try serverless instances, and uses cargo-nextest for CI testing across PG versions. For ML teams, the key implications are easier multi-tenant isolation, faster safe rollouts and migrations, and operational savings from separating scalable storage from compute — useful when running many concurrent agents or experimentation pipelines that need durable yet cheap database snapshots.