October 2026 Summaries

Supabase is joining with Turso to develop database infrastructure for AI agents, which are increasingly creating large numbers of applications, prototypes, and associated databases. Existing products will continue independently, with Supabase focused on Postgres for scalable production workloads and Turso focused on SQLite for lightweight, on-demand databases. Turso’s Rust-based SQLite cloud architecture can operate millions of databases from a single server by loading and suspending them as needed, making low-cost per-agent databases more practical. The companies plan to provide a consistent developer experience from experimentation through production while maintaining their shared commitments to open source and developer-focused infrastructure. Turso’s team, including cofounders Glauber Costa and Pekka Enberg, will join Supabase, with Costa leading the agentic infrastructure initiative.

Supabase announced agent-focused tools that move backend development workflows into repositories, enabling coding agents to edit SQL schemas and project configuration, generate migrations through the new pg-delta diff engine, synchronize dashboard changes with code, and test changes locally. Local Supabase can now run as native processes without Docker and supports separate stacks for multiple directories or worktrees, though the feature is currently in alpha. The company also introduced Supabase Compute, a private-alpha platform for deploying persistent web services, agents, background jobs, and sandboxes with configurable resources, Linux environments, public or private networking, CLI and API management, and integration with existing Supabase services. In addition, developers can add customizable MCP servers to their applications so users’ AI agents can act on their behalf while authenticating normally and remaining constrained by Row Level Security policies; Supabase provides an Edge Function-based implementation, supporting libraries, middleware, and a headless agent-first application template.

Supabase announced tools designed to help coding agents investigate production issues, test fixes in isolated environments, and operate with more controlled permissions. New capabilities include raw project-log access through an MCP SQL tool, expanded Advisor health checks for abnormal error rates across core services, and a Database Connections view for identifying long-running, blocked, idle, or locking queries and managing their sessions. Supabase also introduced Notebooks, which combine reusable database or log queries with notes and can be created, synchronized, and run by developers or agents through the dashboard and CLI. Updated agent skills provide guidance on monitoring health, security, performance, and resource signals, while improvements to project code representation and local runtimes support testing changes before pull requests. Security controls include enterprise-managed MCP authentication via Okta, scoped read-only-by-default personal access tokens, and MCP confirmation prompts for paid resources or potentially destructive SQL actions. Additionally, Supabase Pipelines, now in public alpha for paid plans, replicate Postgres data to analytical destinations such as BigQuery, ClickHouse, DuckLake, and Snowflake in near real time to reduce analytics workloads on primary databases.

Supabase announced three initiatives aimed at addressing different Postgres scaling challenges: Multigres, OrioleDB, and dbarena. Multigres is an open-source Postgres operating system, built by the Vitess team, that combines scalable connection pooling with automated multi-node failover using synchronous replication and consensus to preserve committed writes, and it is currently available through a private alpha. OrioleDB is an open-source Postgres storage engine in public beta that replaces traditional heap storage with an undo-log approach intended to eliminate table bloat, VACUUM overhead, and transaction ID wraparound while improving concurrency and throughput; Supabase reports up to 1.8 times higher throughput in its published benchmark. dbarena is a public benchmarking site that compares managed Postgres offerings from Supabase, Amazon RDS, and Google Cloud SQL using reproducible methods, raw data, cost-normalized performance measures, latency, and infrastructure specifications.

Supabase announced a code-first platform strategy at Supabase Select, aiming to let developers and coding agents manage application backends, monitoring, and scaling through repository-based schemas and configuration. New capabilities include native local development without Docker, Declarative Schemas 2.0 with generated migrations, private-alpha Supabase Compute for long-running services, and tools for embedding an MCP server in applications so users’ agents can access data under existing authentication and row-level security rules. Operational additions include agent prompts for read-only health, security, performance, and resource checks; expanded Advisors; database connection visibility and controls; direct log querying for agents; a unified Explorer workspace and shareable Notebooks; enterprise-managed MCP authentication, scoped access tokens, and confirmation prompts for potentially costly or destructive agent actions. Supabase also introduced near-real-time Postgres data pipelines to analytics platforms, while its scaling roadmap includes private-alpha Multigres for multi-node PostgreSQL high availability, public-beta OrioleDB as a storage engine intended to reduce bloat and VACUUM requirements, and dbarena, an open benchmarking initiative that publishes methods, data, and cost-performance results.