Cinch is in public alpha.
Multi-Tenant SaaS
Give every tenant their own database. Not a row in someone else's cluster, but a separate, isolated store. The 80% that are idle cost next to nothing.
Shared clusters, shared problems
500 tenants on one Postgres cluster. One of them runs a bad query and everyone gets slow. Your RLS policies are three pages of SQL that nobody wants to audit. Enterprise Customer #47 asks for a data export and you're writing a one-off script to pull their rows from 30 shared tables.
The obvious fix — a dedicated database per tenant — is wildly expensive. RDS at ~$105/instance gets you to $105K/mo at 1,000 tenants. And 800 of those instances are sitting idle.
One database per tenant
Cinch makes per-tenant databases cheap enough to actually do it. Active tenants run on NVMe at $0.50/GB. Idle tenants archive to cloud storage at $0.20/GB. No shared state anywhere.
- → Create a tenant database via API — milliseconds, not minutes
- → Each database is physically separate. No key prefix hacks. No RLS to maintain.
- → Idle tenants auto-archive. 80% of your fleet costs almost nothing.
- → Data export is downloading a file. Deletion is deleting one. Compliance gets simple.
- → Noisy neighbors can't happen — separate storage, separate I/O paths.
The math
| Tenants | Shared RDS | Cinch |
|---|---|---|
| 10 | ~$70/mo | ~$3/mo |
| 100 | ~$300/mo | ~$26/mo |
| 1,000 | ~$2,000/mo | ~$260/mo |
Assumes 1GB per tenant, 80% idle. Active = $0.50/GB NVMe. Idle = $0.20/GB cloud storage.