Turso is the SQLite-at-the-edge bet, fork of libSQL with embedded replicas, a Rust rewrite in active development, and pricing that makes database-per-tenant economic. Reads from a local replica are sub-millisecond; writes go to the primary and fan out. The trade is real: SQLite's write throughput ceiling, no shared transactions across nodes, and a smaller ecosystem than Postgres. For edge-first apps, offline-capable mobile clients, and per-tenant database architectures, Turso is the right shape.
- Embedded replicas, sub-millisecond reads on the edge or in-process
- 100 databases on the free tier, database-per-tenant is economic
- libSQL fork is open source, Rust rewrite in active development
- Native React Native bindings (2026) + BYOK encryption
- First-class fit with Hono, Drizzle, and edge-deployed frameworks
- SQLite write ceiling, not for write-heavy OLTP
- Concurrent writes still in private beta
- Smaller ecosystem than Postgres / MongoDB, fewer ORMs, tools, integrations
- Single primary writer per database, replication is read-only
- Newer than the alternatives, fewer war-stories at scale
Turso is a SQLite-based database platform built around embedded replicas: the database lives at the edge (or in your application process) for sub-millisecond reads, with writes routed to a primary. It started as a managed fork of libSQL (Turso’s open- source extension of SQLite), and a full Rust rewrite of SQLite is in active development, already a year in as of early 2026.
Where it fits
Turso is the right pick for edge-first apps where read latency needs to be sub-millisecond everywhere, global SaaS, dashboards, docs sites, anything where the database serving from one region adds noticeable lag. Embedded replicas let mobile apps run a local SQLite that syncs with the primary, which makes offline-first sync genuinely simple.
The pricing model also makes database-per-tenant economic in a way that Postgres doesn’t, the free tier alone supports 100 databases, and even the $25/month Scaler tier supports unlimited. For multi-tenant apps where isolation matters and tenants are many, Turso is uniquely shaped.
Avoid Turso for write-heavy OLTP, complex transactional workloads spanning many tables, or anything that would otherwise push Postgres to its limits, SQLite’s strengths aren’t there.
Pricing in practice
Free includes 100 databases, 5 GB storage, 500M monthly row reads, and 10M writes, generously sized for prototypes and small production apps. Developer is $4.99/month with unlimited databases, 9 GB storage, and 2.5B row reads. Scaler is $24.92/month (or $299/year) with 24 GB storage, 100B reads, team support, and 14-day audit logs. Pro is $416/month with 50 GB storage, SOC2, HIPAA, BYOK, and 30-day audit logs.
Overage pricing is itemized per resource, storage at $0.50-$0.75/GB, reads at $0.75-$1/billion, writes at $0.75-$1/million. The free tier is unusually generous and covers more real apps than typical free tiers.
How it compares
Supabase, Managed Postgres with the full backend platform. Pick when you need a real Postgres with auth/storage/realtime bundled, not an edge-first read store.
Neon, Serverless Postgres with branching and scale-to-zero. Pick when you want Postgres semantics and elastic compute, not edge replicas.
PostgreSQL, The default relational database. Pick when write throughput, joins, and a deep ecosystem matter more than edge reads.
What changed recently
Turso v0.6.0 shipped May 12 2026, alongside the launch of the Turso Startup Program. Turso Sync landed in April 2026 with a ground-up rewrite of the sync engine. The next-generation Turso Cloud entered private beta April 8 2026 with concurrent writes (historically a SQLite limitation), unlimited active databases for all users, BYOK encryption, and React Native bindings. The Rust rewrite of SQLite is “a year in” as of March 2026 and continues to track toward replacing the C engine entirely.
Sources
- Turso Blog, turso.tech, May 2026
- Turso Pricing, turso.tech
- Turso Sync, turso.tech, April 24 2026
- Turso Database (Rust rewrite), github.com