The SQLite rewrite company that bet databases would become files again — acquired by Supabase in 2026
HQ: San Francisco, California, USA (technically SF-based; globally distributed remote team across four continents) | Founded: 2021 | Employees: ~22 (PitchBook and Tracxn list 22 employees in 2026; Crunchbase lists 11–50) | Stage: Acquired by Supabase (announced October 2, 2026, terms undisclosed); previously seed-stage private | Website: https://turso.tech
Turso is a database company that spent its life making SQLite viable for the cloud and the agent era — first by forking it, then by rewriting it from scratch. It was founded in 2021 in San Francisco as ChiselStrike, Inc. by three ex-ScyllaDB engineers: Glauber Costa (CEO), Dejan Mircevski and Pekka Enberg. Norwest Venture Partners led a $7 million seed in July 2022, with Blumberg, First Star, Essence VC, Mango Capital and Netlify's Jamstack Innovation Fund participating.
The company pivoted twice. ChiselStrike's TypeScript backend platform gave way in late 2022 to libSQL, an open-contribution fork of SQLite, and in January 2023 to ChiselStrike Turso, an edge database that replicated SQLite data to dozens of regions over HTTP. The company then renamed itself Turso. In December 2024 it announced Limbo — a complete rewrite of SQLite in Rust — and went all-in in January 2025, deprioritizing the older replication platform. The engine shipped as Turso in alpha in July 2025 and reached version 0.8 in September 2026: SQLite-file-compatible, with MVCC concurrent writes, async I/O, vector and full-text search and sync — MIT-licensed, with 260+ contributors.
Turso Cloud meters storage, rows read, rows written and syncs from a free tier to enterprise and BYOC, running databases as files on diskless, WAL-on-S3 storage so idle fleets stay cheap. Named customers include Superhuman, Mastra, CTO.new, Sauna.ai and Adaptive (2M+ databases), plus embedders like Astro and Fermyon. In July 2026 Turso said it was building a Postgres-compatible frontend — the 'LLVM of databases' positioning. On October 2, 2026, Supabase announced it was acquiring Turso for an undisclosed sum; Costa joined Supabase as Head of Agentic Services and the platform continues to operate.
Turso's thesis is that the unit of database deployment is changing: instead of one big Postgres per application, the agent era needs millions of tiny, isolated, instantly available databases — one per agent, tenant or user — and the economics only work if a database is a file rather than a process. That came out of the founders' ScyllaDB and Linux-kernel backgrounds: the most deployed database in the world cannot write concurrently and does not accept outside contributions.
What makes Turso unusual is the path taken to exploit it. Rather than build a proprietary engine, the company first forked SQLite as libSQL in late 2022, then — concluding a fork could never lift SQLite's core limitations — rewrote the engine from scratch in Rust (announced December 2024 as Limbo, all-in January 2025, first alpha July 2025, version 0.8 in September 2026). The result adds MVCC concurrent writes, async I/O, vector and full-text search and bidirectional sync while keeping SQLite's file format. In July 2026 it added a Postgres frontend compiled onto the same bytecode VM — the 'LLVM of databases' bet.
The financial context matters as much as the technology. Turso raised roughly $7 million in disclosed venture funding — a rounding error next to Supabase ($1B+), Neon ($129.6M, sold to Databricks for about $1B) or PlanetScale ($105M), and capital intensity kills database startups: Fauna shut its service in May 2025 after failing to raise. Turso's answers were extreme capital efficiency, 260+ contributors and agent-native customers (Superhuman, Mastra, CTO.new, Adaptive with 2M+ databases) — but ultimately a sale. Supabase acquired Turso on October 2, 2026 for an undisclosed price and made Costa head of agentic services, turning a direct competitor into the engine room of its agent strategy. The open questions: whether SQLite-compatibility ceilings matter as agents outgrow lightweight stores, and whether Turso's architecture survives as a product inside a Postgres company.
Browse: Home | Blog | About | All Startups A-Z | View full profile on StartupWiki