multi-tenant-postgres

Row-level security (RLS) enforces per-tenant data access at the database layer within a single PostgreSQL instance. Set the current tenant context (project_id) on each request, apply RLS policies tied to current_setting, and PostgreSQL automatically filters rows by tenant—eliminating the complexity of managing separate databases or shards.

Yes. RLS policies isolate data for organizations, projects, and agents in a single PostgreSQL database. Define policies that reference the current tenant context, and all queries automatically respect tenant boundaries—reducing operational overhead while maintaining strong isolation.

Use versioned migrations to create your core schema (organizations, projects, agents) and define RLS policies that enforce tenant isolation. Run migrations during deployment to initialize the schema and activate policies; pair migrations with sqlc for type-safe, tenant-aware queries.

Use PgBouncer to pool connections and manage tenant context efficiently across requests. Set the current_setting on each connection to establish the tenant scope; PgBouncer forwards requests to PostgreSQL while RLS policies enforce isolation, reducing per-connection overhead at scale.

sqlc generates type-safe queries from SQL, reducing errors when building tenant-aware data layers. While not required, it integrates well with RLS policies and migrations to ensure queries respect tenant isolation and maintain consistency across your application and agent stacks.