Tested backups

What your agent writes

Your agent writes it in while.yaml, at the root of your app:

databases:
  - name: main
    backup: { schedule: "0 3 * * *", test_restore: true }

When to use it

They are always there for a database, without declaring anything: go back to a backup or to a precise point in time, over 14 days.

How it works

Each database is backed up continuously and every day, outside the cluster, and the daily backup of a production database is tested by a real restore. Your agent puts a database back to a backup or to a precise point in time within the last 14 days.

What to know

Each database archives its log (WAL) continuously and makes a full backup per day, encrypted, outside the cluster. schedule sets the time of this backup, in UTC. Backups are kept 14 days, for all databases: retention_days is read, but not followed.

The daily backup of a production database, and any backup started by the backup_database tool, is followed by a test: a throwaway database is born from it, and its tables are read back. The result is shown, never assumed.

restore_backup, approved by a person, puts a database back to a backup, or to a precise point in time within these 14 days, starting from the last backup finished before that point. The restored database is born new, and takes the old one’s place after a safety backup. Backups have a single copy, in a single region.

Pricing

These prices are the same on every plan, excluding VAT:

Your agent’s MCP tools

Your agent handles it with the get_environment, backup_database and restore_backup tools.

Guides

• Restore your PostgreSQL database to a precise moment, or from a pg_dump, from your AI agent