Serverless eterDB
Serverless Postgres for a world where the primary user of your database is an agent, not a person clicking through a dashboard. The CLI is the interface. Undo is the safety net. Everything else — scale-to-zero, branching, a pretty dashboard — is table stakes, and Serverless eterDB has that too.
Built for agents to run your database, not just use it
Every serverless Postgres on the market was designed for a person clicking through a dashboard, with an API bolted on as an afterthought. That was a reasonable bet two years ago. It isn't now: agents provision databases, run migrations, seed data, and clean up tables at 2 AM, unattended. A GUI-first product with an API as a courtesy can't be the primary interface for that. Serverless eterDB inverts the order — the CLI comes first, and recovering from whatever the agent just broke is the same command it already knows, not a ticket to support or a restore from last night's backup.
inventoryliveThat revert is eterDB's surgical undo, running underneath Serverless eterDB — reverse a transaction down to the exact rows it touched, without a full restore or downtime. It's the one thing a dashboard-first competitor can't retrofit in. See how it works for the underlying mechanics.
Table stakes, handled
Scale-to-zero, copy-on-write branching, a management app, usage-based pricing — Serverless eterDB has all of it, because a serverless Postgres that can't do these isn't a serious option in 2026. They're the entry price, not the pitch:
eterDB is open source. Serverless eterDB is us running it for you.
The undo engine underneath Serverless eterDB — eterDB — is open source and self-hostable. Clone it, run it on a Linux host you control, and your agent gets the same CLI and the same surgical undo, no waitlist required. Serverless eterDB is us running that engine for you: serverless, branching, zero ops — the table stakes handled, so the agent-first part is all you have to think about.