What is Supabase? (the database your AI keeps choosing)

What it actually is

Supabase is a hosted backend built on top of Postgres, a real, mature relational database, with a bundle of extra services wrapped around it: user login (auth), instant APIs so your app can talk to the database over the web, file storage, and realtime updates that push changes to connected users as they happen. You create a project on supabase.com, and within a couple of minutes you have a live database with a dashboard, not a folder on your laptop.

Think of it like renting a fully fitted workshop instead of building one from raw materials. The workbench itself is Postgres, sturdy, decades-proven furniture. Supabase adds the electrical wiring (auth), a loading dock with a delivery log (storage and APIs), and an intercom system that announces the moment something changes on the bench (realtime). You didn’t have to wire any of that yourself.

Why your AI just did this

When your agent needs your app to remember something between visits, log a user in, or store an uploaded file, it needs an actual backend, not just the frontend code that renders your screen. Supabase is the default pick across most AI builders (Lovable, Bolt, and others) because it packages a production-grade database with auth and storage behind one API key, so the agent can say “save this form entry” or “check if this user is logged in” without standing up its own server.

On Lovable specifically, the backend that’s there from the start is called Lovable Cloud, and it is a Supabase project, just one that Lovable owns and manages for you. You won’t find it if you log into supabase.com yourself. “Connecting Supabase” instead means linking a project under your own Supabase account, which your agent can then read and modify directly, including through the official Supabase MCP server, a connector that lets Claude Code list your tables, run SQL, and apply migrations on your behalf.

When you’ll run into it

• Your agent says something like “I’ve created a profiles table in Supabase” or “you’ll need to connect a Supabase project first”
• Inside Lovable, you see two separate options: Lovable Cloud (already active, no extra login) and Connect Supabase (opens a real Supabase account flow)
• An error like new row violates row-level security policy, that’s Postgres’ Row Level Security blocking a query because no rule yet allows it
• You ask to see your actual data and get pointed to a dashboard at supabase.com, not a file on your machine
• Your agent runs a command through the Supabase MCP server, listing tables or executing SQL directly against your project

What to check

• Whether you’re on Lovable Cloud (managed by Lovable, nothing to log into) or your own Supabase project (you have a supabase.com account with a project URL and API keys)
• Row Level Security is turned on for any table holding real user data. Tables your agent creates through SQL or a migration (the usual path) start with RLS off, which means anyone with your public key can read or write them until you add a policy
• Your Supabase project URL and keys live in environment variables, never typed directly into a component
• If your agent connected through the Supabase MCP server, confirm it’s pointed at a development project, not the one your real users touch
• That you personally can log into supabase.com and see your own tables. If you can’t, you don’t actually control that data

Say it like a dev

Instead of: “the AI’s database” Say: “the Supabase project” or “our Postgres database”

Instead of: “why can’t I see my data anywhere” Say: “is this on Lovable Cloud or our own Supabase project, and do I have the dashboard login?”

Instead of: “lock down the database” Say: “did we turn on Row Level Security policies for this table?”