Supabase, the open-source backend platform, is reshaping its database and developer tools for a future in which AI agents build and run much of the software. In a 77-minute Latent Space interview published October 28, co-founders Paul Copplestone and Ant Wilson described how the company is adapting its product for that shift. Host swyx said an AI-tracking test he ran showed Supabase as the first choice for application backends across nearly every frontier model he checked, and it was in contention for managed databases. The company's recent announcements focus on removing setup work that agents and developers both encounter. A new alpha runtime lets developers run a local Supabase backend without Docker. Declarative schemas, kept in Git, give an agent a description of the desired database state that it can read, edit and fork, rather than a long chain of migrations. Supabase Compute, now in private alpha, runs containers beside the database for jobs such as chunking documents and generating embeddings. The company also says its Operate tools, which include logs, checks and connection data, give an agent the context it needs to understand failures. The founders said the company launches about five million databases a month and is preparing for a future in which agents push that number far higher. Existing customers are also running much larger workloads, which the company is addressing through OrioleDB, the Multigres high-availability Postgres project in private alpha, and its acquisition of Turso. Copplestone and Ant said that as AI-driven usage grew, activation and conversion improved, because an agent can take a user from signup to a working app in a single flow. These are company statements from the interview and have not been independently verified. For Australian businesses and professionals, the practical impact is limited for now. Most of the features discussed are in alpha or private alpha, and the interview did not cover pricing changes. Teams already building AI-assisted applications on Supabase could test the local runtime or keep schemas in version control, which is standard practice. Any application that stores personal information remains subject to the Privacy Act 1988 regardless of which backend an agent chooses, so data location, access controls and vendor terms still need review before production use.