SQLite is the most-deployed database on earth, but it was written in C decades ago. Turso is a ground-up rewrite in Rust that keeps SQLite compatibility while adding the things modern AI apps actually want — and it’s surging on GitHub as developers look for a local data layer built for agents.
## What Turso is
Turso runs in-process: databases are files, always on, with no cold start or server to wake. The rewrite is async-first, adds incoming concurrent writes, and runs in the browser via WASM. The AI-relevant piece is native vector search using the DiskANN algorithm, with float32 down to aggressive 1-bit quantization (32x compression) — so retrieval-augmented generation can live inside the same embedded database as your app data, no separate vector store.
## A database per agent
Turso leans into a “many-database” model: every agent, user, or tenant gets its own database. Teams are already using its cloud to spin up large numbers of ephemeral databases for AI agents, branching and rolling them back with a single API call. The Rust engine is still pre-production, while the C-based libSQL fork remains the production-ready path today.

Leave a comment