What is Silos?
A serverless, edge-native PostgreSQL platform that runs a real Postgres engine compiled to WebAssembly.
Silos is a serverless PostgreSQL platform. You get real PostgreSQL — not a subset, not a wire-compatible reimplementation — provisioned on demand, billed by what you use, and able to run close to your users at the edge.
What makes Silos different is how it runs Postgres. Instead of keeping a database server running on a virtual machine, Silos compiles PostgreSQL to WebAssembly and runs it inside a sandboxed WASM runtime. Your database's durable state — schema, rows, indexes — lives as compressed snapshots and write-ahead-log (WAL) segments in object storage. A database is "just data" until something queries it, at which point a runtime materializes it and starts serving.
Why that matters
This architecture unlocks a set of properties that are awkward or expensive with a traditional always-on Postgres server:
Instant provisioning
A new database is a new set of metadata and storage — created in seconds, not minutes.
Scale to zero
When nothing is querying, the compute spins down. You don't pay for idle databases.
Branch like Git
Copy-on-write snapshot forks let you branch a database for a preview, a test, or an experiment.
Edge replicas
On-demand replicas materialize near your users so reads happen locally.
Same engine in the browser
PGLite runs the same Postgres engine in WASM client-side for local-first dev.
Agent-native
Create and destroy databases through the API as fast as an automated workflow needs.
Real PostgreSQL
Because the engine is genuinely PostgreSQL, you keep the things you rely on:
transactions and MVCC, PL/pgSQL, triggers, views, CTEs and window functions,
JSON/JSONB, full-text search, and the standard SQL surface. And because Silos
speaks the standard PostgreSQL wire protocol, your existing clients, drivers,
and ORMs connect the same way they would to any Postgres.
psql "postgres://user:password@db-xxxx.silos.sh/main?sslmode=require"See PostgreSQL compatibility for the details, and Silos vs. traditional Postgres for how the model compares.
Where Silos is today
Silos is in active development. The architecture is real and substantial — a WASM Postgres runtime, a wire-protocol proxy, snapshot/WAL storage, branching, a metering system, a CLI, and a full web console. Some headline capabilities (global WAL streaming, edge replica materialization, multi-region production) are still rolling out, and the docs label those clearly. See When to use Silos for an honest fit assessment and the roadmap for what's next.
Throughout these docs, anything described as a target is a design goal we're building toward, not a measured guarantee. Anything marked preview or rolling out is implemented but still maturing.