Cinch is in public alpha.
Cinch
File-backed databases with cloud storage economics.
Cinch is a database platform where every database is a file. Multi-tenant infrastructure underneath, fully isolated workspaces on top. Create a database in milliseconds via API, use it with protocols you already know, and pay almost nothing when it's idle.
We build on embedded engines and NVMe with tiered cloud storage. No server processes. Redis, Bolt/Cypher, and SQLite — without making you run the servers behind them.
What we believe
Databases shouldn't need servers.
The pattern is already proven. S3 runs compute over private, multi-tenant file-backed data at massive scale. The same abstraction applies to databases.
Idle data shouldn't cost active prices.
Most databases are idle most of the time. Paying RAM prices for cold data is a structural problem, not an optimization opportunity. NVMe and cloud storage change the economics.
Isolation should be the default.
Key prefixes and row-level security are workarounds for a missing primitive. Every workload — every tenant, agent, branch, device — should get its own database. The cost model just needs to support it.
The database lifecycle should be API-driven.
Create, use, observe, branch, migrate, sleep, wake, delete. Every operation programmable. No dashboards required.
Why now
Three shifts are converging. NVMe closed the performance gap with RAM — 100x cheaper, and over the network the difference nearly disappears. RAM is getting dramatically more expensive, with no relief until 2027+. And agents are creating databases at a pace that server economics can't absorb.
The managed database market is $150B+ today. Agents will multiply the number of database instances per application by 100-1000x. Current infrastructure wasn't built for that.
Based in New York City.