Neon

Promptfree, for everyone, and the only version there is

Copy the prompt for Claude CodeCopy the prompt for CodexCopy the prompt for Cursor

Copy

Do not build this. Branching and scale-to-zero storage is the product, and it is a storage engine rather than a configuration.

Separating storage from compute so a database branches instantly and restores to any point in time means a copy-on-write layer underneath Postgres. That is systems work, not a script.

WHEN TO KEEP PAYING
- Preview environments per pull request with real data shapes, which is genuinely useful and genuinely hard otherwise
- Traffic that is spiky or occasional, where paying nothing while idle actually matters

WHEN NOT TO
- Steady traffic on one application. Then Postgres on a machine you rent is cheaper, faster and has no cold start
- Small enough that SQLite on the application's own machine is the honest answer — reads become microseconds and there is no network at all

WHAT REPLACES EACH FEATURE, CHEAPLY
- **Branching** becomes a copy of the database restored into a preview environment, with the data anonymised on the way. Write the anonymiser as part of the branch script, not as something somebody remembers. For SQLite that is one `VACUUM INTO`; for Postgres a dump and restore, and a scheduled cleanup of old branches
- **Point-in-time restore** becomes continuous replication of the write-ahead log to object storage. Litestream for SQLite, or the archive facility for Postgres — recovery point in seconds rather than a night, and it is the feature this product is most admired for
- **Scale to zero** becomes one small always-on instance for a few euros, which has no cold start at all — usually the better trade
- A connection pooler in front, because a serverless application opening a connection per request is the failure this category exists to paper over

THE PART THAT MATTERS MORE THAN THE PLATFORM
Rehearse a restore onto a clean machine and write down how long it took. That number is your recovery time, and almost nobody knows theirs. Migrations backwards-compatible across a deploy: add a column, deploy, use it, remove the old one later — never in one step.

THE ONE-LINE VERSION
Buy it for previews and spiky traffic. Otherwise set up replication, script the branch, and rehearse the restore.