Starter architectures - Korve

Korve’s starter guides are transparent architecture specifications, not opaque one-click templates. Use the smallest matching shape in your own repository, review its manifest plan, and deploy only after the declared resources match your application.

Full-stack SaaS

A Bun application shape with App Auth, tenant-scoped Postgres rows, object storage, a recovery policy, and a strict browser/server credential boundary. The design accounts for one-time database credentials and explicit environment configuration.

Agent-operated internal tool

An organization-only application with a stable deployment identifier, exact health evidence, daily maintenance analysis, and approval-gated operator actions. No organization credential is embedded in the application.

Queue and cron worker

A signed push consumer with durable replay protection, stable idempotency, retries, dead-letter handling, and a timezone-aware cron that publishes into the work queue.

Safe workflow

korve manifest plan -f korve.toml
korve manifest apply -f korve.toml
korve deploy .

Run your repository’s own tests and build before this workflow. Optional resources are provisioned only when declared; ordinary deployment does not silently create a database, cache, queue, or storage bucket. Database connection strings and other shown-once values belong in Korve environment variables, never in a manifest or repository.