October 2026 Summaries

7 posts from Turso

Filter
Month: Year:
Post Summaries Back to Blog
Turso Cloud has removed active-database limits from its paid plans, allowing unlimited databases that cost only for storage while idle, with the Developer plan starting at $4.99 per month and including storage and row-operation allowances. The change makes database-per-tenant, per-agent, per-experiment, per-user, and per-preview-environment architectures more practical by reducing the former cost penalty for large numbers of lightly used databases. Turso stores databases as object-storage-backed file segments rather than continuously running server processes, so idle databases require no compute resources or cold starts. Separate databases can provide isolation by construction, simplify tenant-specific backup, restoration, deletion, and compliance discussions, and permit independently tailored schemas, while provisioning can be handled through API calls, seeded templates, scoped tokens, and a routing layer. Shared schemas remain appropriate for early-stage products, similar small tenant workloads, central cross-tenant analytics, or existing well-tested row-level security systems, and hybrid models can isolate only enterprise or regulated customers. Operating many databases still requires solutions for cross-tenant analytics, fleet-wide migrations, request routing, monitoring, and storage management, while free plans remain limited to 100 databases.
Oct 30, 2026 2,096 words in the original blog post.
Turso’s MVCC mode redesigns SQLite-compatible write handling to support concurrent writers while preserving durability and recovery, tracing a row’s lifecycle from SQL compilation into SQLite VDBE bytecode through insertion, commit, checkpointing, and deletion. New rows first enter an in-memory, lock-free SkipMap as version chains marked with transaction IDs, becoming visible to other transactions only after conflict validation, dependency checks, logical-log writing, and fsync establish a durable commit timestamp. The logical log enables crash recovery by replaying committed operations into memory, while checkpoints periodically transfer eligible in-memory data through SQLite’s WAL into the B-tree database file, using either blocking truncate checkpoints or an experimental less-blocking passive mode. Reads merge data from the in-memory MVCC store and persistent B-tree using snapshot isolation, giving newer in-memory versions precedence, while deletes mark a row version’s end timestamp and detect competing writes as conflicts. The implementation aims to overcome SQLite WAL mode’s single-writer limitation, particularly for Turso’s small databases serving parallel agent workloads, while retaining SQLite file compatibility through the `PRAGMA journal_mode = mvcc` option.
Oct 19, 2026 3,408 words in the original blog post.
Multi-tenant SaaS applications commonly use a shared schema with tenant IDs and row-level security, separate schemas on one server, or a dedicated database for each tenant, with each model balancing simplicity, cost, isolation, and operational burden differently. Shared schemas are efficient for similar low-regulation workloads and centralized analytics but rely on correctly enforced logical isolation, while schema-per-tenant adds migration and tooling complexity without fully separating infrastructure. Database-per-tenant provides structural isolation, simpler tenant-specific backup, deletion, exports, and regional placement, making it attractive for regulated industries and enterprise security reviews, though it requires robust request routing, monitoring, and separate cross-tenant analytics pipelines. The article argues that SQLite-based platforms such as Turso reduce the historical expense of operating many databases by making idle databases storage-only and automating provisioning, token scoping, and teardown, citing deployments with thousands or millions of isolated databases. It recommends shared schemas with tested RLS for early or less regulated products, dedicated databases for customers requiring stronger isolation or data controls, and hybrid approaches for serving both smaller and enterprise tenants.
Oct 14, 2026 1,851 words in the original blog post.
Turso has launched a plugin in the Cursor Marketplace that lets coding agents create, branch, query, and manage Turso databases directly within the editor. Installation requires the `/add-plugin turso` command and browser-based OAuth sign-in, avoiding the need to create or store API tokens in project configuration files. Through the plugin, agents can manage cloud databases, test migrations against database branches, run SQL for inspection and debugging, administer organizations and groups, and review usage analytics. It also includes a Turso skill that provides agents with guidance on SQLite, libSQL, Turso tooling, drivers, and technical tradeoffs. Built on Turso’s hosted MCP server, the integration can also be used with Claude Code and Codex, while its source code is available on GitHub.
Oct 08, 2026 331 words in the original blog post.
Toasty, an asynchronous Rust ORM created by Tokio author Carl Lerche, now fully supports Turso databases in three deployment modes: local files with concurrent writes, fully remote serverless databases accessed over HTTP, and local-first replicas synchronized with Turso Cloud. Designed to bring Rails-like simplicity and ergonomics to Rust development, Toasty lets developers use the same models and query APIs across these modes, changing only the database driver configuration. Local Turso databases can use automatic `BEGIN CONCURRENT` transactions for parallel writes, while remote deployments can improve efficiency by batching independent statements into a single HTTP request. In synchronized deployments, developers can explicitly pull remote changes and push local updates at intervals or key durability points. The integration aims to let applications start with a zero-setup local SQLite-compatible database and later move to cloud, synchronization, or serverless architectures without rewriting their data layer.
Oct 07, 2026 1,276 words in the original blog post.
Supabase is acquiring Turso, an open-source SQLite-based database platform designed to support massive numbers of lightweight, isolated databases for AI agents and other server workloads. Turso says it addressed SQLite’s traditional single-writer and server-scale limitations through a redesigned architecture using concurrent writes and diskless WAL-on-S3 storage, enabling inexpensive, on-demand provisioning in its cloud or through bring-your-own-cloud deployments. The companies plan to combine Turso’s “many database” approach with Supabase’s Postgres platform, authentication, storage, and ecosystem, offering users a path from per-agent SQLite databases to standard Postgres and potentially petabyte-scale Multigres deployments. Turso will continue operating, its databases and APIs will remain available, and its database software will stay open source, while founder Glauber Costa becomes Supabase’s Head of Agentic Services alongside cofounder Pekka Enberg and the rest of the Turso team.
Oct 02, 2026 660 words in the original blog post.
Turso Cloud has expanded to four new AWS regions in Sydney, São Paulo, Central Canada, and Stockholm, bringing its total footprint to ten regions across four continents. The expansion enables developers in Australia, Brazil, Canada, and Sweden to place databases nearer to applications and users, reducing round-trip latency from more than 100 milliseconds in some cross-region cases to roughly 1 millisecond within a region while supporting in-country data residency needs. Turso’s architecture uses AWS S3 as its underlying storage foundation, writing transaction logs to the low-latency S3 Express One Zone service before asynchronously compacting data into standard S3, a design intended to scale to billions of databases and operate within customer cloud accounts. The new availability follows AWS’s September 2026 expansion of S3 Express One Zone into additional regions, with Turso also considering future support for Singapore, Paris, Seoul, and Northern California.
Oct 01, 2026 584 words in the original blog post.