One architecture.
Every database you'll ever need.
Turso is built on SQLite and rewritten in Rust for a world that needs many databases, not one large one. The same engine runs on your laptop, in the browser, on a device, and across the cloud. Provision one database per agent, user, or tenant, and treat them as a fleet.
Three ways to run Turso
Turso is one engine with three ways to run it. The open-source engine embeds directly in your application. Turso Cloud hosts your databases as a managed service. Embedded replication keeps a cloud database and an on-device copy in sync. You write against the same SQLite-compatible interface in every case, so code moves from prototype to production without a rewrite.
Open-source Turso
A SQLite-compatible engine you can embed anywhere. Offline, in the browser through WebAssembly, or on-device. MIT licensed and built in the open.
Explore open-source TursoTurso Cloud
Unlimited databases that never sleep, accessed over HTTP or replicated to your own devices. Provisioning, branching, observability, and per-database encryption as a managed service.
See how Turso Cloud worksEmbedded replication and sync
Run the same database on a device and in the cloud. Read and write locally with zero network latency, then sync when the device reconnects.
How embedded replication worksWhat the engine adds to SQLite
Turso is a ground-up rewrite of SQLite in Rust, intended as a drop-in replacement. It keeps the SQLite SQL dialect, file format, and C API, then adds the capabilities modern workloads need.
Concurrent writes
Multiple writers make progress at once through multi-version concurrency control, using BEGIN CONCURRENT. Classical SQLite serializes writes.
Native vector search
Similarity search is built into the database, with no extensions, for RAG and on-device AI.
Async I/O
An async-first model built on primitives like Linux io_uring, suited to serverless and edge runtimes.
Browser and WebAssembly
Run Turso client-side with WASM and OPFS persistence.
HTTP and WebSocket access
Reach remote databases without holding open a TCP connection.
Change data capture and FTS
Track changes in real time and search text natively.
Built for the many-database architecture
Traditional databases were designed around a single shared instance. Scaling meant adding CPU, memory, or replicas to that one instance. Turso flips the model. Because a database is a file rather than a long-running process, you can create thousands or millions of them, and an idle database costs only its storage footprint. There is no process to wake, no cold start, and nothing to scale to zero.
