A backend toolkit you run on your own machine and ship to your own server. It reads your schema before every deploy, and stops when it finds a table the internet can read. Out today.
$ cresco deploy --server web-1 Access control critical users is readable by anyone, and holds email address critical appointments has no row owner: every signed-in user can read every row Refusing to deploy. Nothing was written. Override with --allow-public if this is deliberate.
Most tools scan a URL after you have already shipped the mistake. CrescoDB owns your schema, so it answers before anything leaves your machine.
Database, REST API, auth with roles, file storage and realtime. SQLite, PostgreSQL or MySQL, with the same schema and the same API on all three, so there is no rewrite at deploy. A built-in API explorer and SQL editor, and scaffolds that write real code into your repo rather than hiding it behind a dependency.
One command connects Claude Code, Cursor or Codex over MCP. Your agent reads the real schema instead of a description you pasted, calls the API it just generated, and can ask whether anything is unsafe to deploy and get an actual answer. A hosted backend can only be guessed at from the outside.
Every platform says "backup succeeded", which only ever means a file got written. Ours are loaded into a throwaway database and counted. A backup that restores empty is reported as a failure, because an empty file is not a backup. You also get told when your app stops answering, which is the failure that looks healthy in every dashboard.
cresco eject writes a folder that runs your project with no CrescoDB account and no dependency on us: your code, your data as plain SQL any database client can read, and the engine packaged inside. Run it on day one, before you have committed anything.
Free forever on your own machine. No card.