Spacetime

Spacetime collapses what normally takes a dozen services into a single system. Instead of wiring together a web server, ORM, relational database, cache, WebSocket layer, and sync protocol, you write your server logic directly inside the database as a module, and Spacetime handles the rest.

It's built for apps and games that need real-time state: multiplayer experiences, collaborative tools, chat applications, anything where clients need to stay in sync without you hand-rolling a sync protocol. Every subscribed client gets updates automatically when state changes, with no polling and no extra infrastructure.

The core model is built around five guarantees:

• Everything is a table. All application state lives in a relational schema with SQL-style queries.
• Everything is persistent. State is persisted by default, including the full row history.
• Everything is real-time. Subscription queries push updates to clients the moment data changes.
• Everything is transactional. Reducers are atomic: they either complete fully or don't run at all, with strongly serializable ACID guarantees.
• Everything is programmable. Server logic runs inside the database as real code in TypeScript, Rust, or C#.

Because application logic runs embedded in the database rather than over a network hop, Spacetime benchmarks significantly faster than conventional stacks like Node.js with Postgres or tools like Convex. The performance gap widens under contention.

The type system is shared across client, server, and database, so there's no translation layer to maintain and no place for types to drift out of sync. Client SDKs are generated automatically from your module.

Deployment is intentionally minimal. Publishing a module makes your database immediately accessible, with no container orchestration, no networking configuration, and no separate auth service to stand up. Auth is built in.

Spacetime suits solo developers and small teams building real-time products who want to move fast without managing distributed infrastructure. It's also a natural fit for AI-assisted development workflows, since the unified type system and single deployment target reduce the surface area where generated code can go wrong.