Supabase Buys Turso to Give AI Agents Their Own Databases

Supabase is acquiring Turso, the company that rewrote SQLite in Rust, to build databases that agents can create as easily as files, CEO Paul Copplestone said in an Oct. 2, 2026 post (Supabase blog, Oct. 2, 2026). Supabase said it already launches more than 1 million databases a week.

What Supabase and Turso said

Copplestone framed the deal around agents. Agents “are spinning up millions of databases” for prototypes, dashboards and apps, he wrote, and that demands cheap, on-demand databases that don’t need a dedicated machine each. SQLite fits small workloads, he said, and Postgres is what you want as an app scales. The goal is the same developer experience from prototype to production.

Turso’s own announcement, from co-founder Glauber Costa the same day, described the architecture: a diskless, WAL-on-S3 design that lets one server manage millions of databases, loading them when needed and suspending them when idle. Costa named Superhuman, Sauna.ai and Mastra as customers.

Both companies made the same commitments. Turso keeps running, and existing databases and APIs continue as they do today. Turso Database stays open source. Supabase will keep building around Postgres while Turso keeps working on SQLite. Costa joins Supabase to lead the agentic infrastructure effort, alongside Pekka Enberg and the rest of the Turso team. Turso said the graduation path runs from SQLite into standard Postgres through Supabase, and to petabyte scale with Multigres, with “deeper integration” promised in the coming months.

How developers reacted

The announcement drew 216 points and 117 comments on Hacker News by Oct. 4. The thread split along a familiar line. One commenter, vmg12, said they had avoided Turso because its future seemed tied to the company’s, and would now choose it for new projects. Another, brainless, said they did not feel good about the deal and wondered whether Turso’s hosting had been earning enough. JoshTriplett hoped Turso would not become “another incredible journey.”

Reliability came up as well. A commenter named deaux said their team left Turso after latency above 100 ms in certain regions that lasted a long time without a status-page notice. Another pointed to ClickBench issues where Turso ran slower than SQLite.

Our take

This is a bet on shape. Supabase’s product is one big Postgres database per project. An agent that needs a scratch database for one task doesn’t want a project, and Turso’s many-small-databases model is built for that case. Buying it is faster than building it.

The test is the graduation path. “Move to Postgres when SQLite isn’t enough” only works if the migration is routine, and SQLite and Postgres differ in types, concurrency and SQL dialect. That move decides whether this is one platform or two products under one logo.

The Hacker News worry about Turso’s independence is the one to watch. The open-source commitment is in writing, and the next few releases of Turso Database will show whether it holds.