When to use Silos
An honest fit assessment — where Silos is a great choice today and where to be careful.
Silos is a young platform with a distinctive model. Here's a straight answer about where it fits today.
Great fits
Serverless & spiky apps
Workloads that are idle much of the time benefit directly from scale-to-zero billing.
Multi-tenant SaaS
A database (or branch) per tenant, created and destroyed on demand, each isolated in its own runtime.
Preview & test environments
Branch a database for every PR, run tests, then throw it away.
Global read-heavy apps
Materialize replicas near users for local reads with per-query consistency.
Agent-native systems
AI agents that provision ephemeral databases through the API and tear them down when done.
Local-first development
Prototype entirely in the browser with PGLite, then deploy the same Postgres.
Be careful with
- Latency-critical first queries. Cold starts add time to the first request after idle. Keep databases warm for hot paths.
- Heavyweight extensions. TimescaleDB, Citus, and similar are not available in the WASM runtime. Check the extensions list first.
- Sustained, write-heavy single-primary workloads. Writes route to the primary; a constant high-write firehose is not the model's sweet spot.
- Strict compliance requirements today. Silos is not yet certified for SOC 2, HIPAA, ISO 27001, or similar. See the compliance roadmap.
Current status
Silos is in active development. Core mechanisms — WASM Postgres, the wire-protocol proxy, snapshot/WAL storage, branching, metering, CLI, and console — are built. Global WAL streaming, edge replica materialization, and multi-region production are rolling out. Treat anything marked target, preview, or rolling out accordingly, and check the roadmap before depending on a capability for production.
Not sure if Silos fits?
Tell us about your workload at silos.sh/contact and we'll give you a straight answer.