An AI builder left the database wide open
The generated schema has no row-level security policies, or has policies that effectively allow anyone with the public anon key to read and sometimes write every table.
Every table locked down, tested
Working with Supabase?
Most Supabase projects that land on our desk were scaffolded by an AI app builder in an afternoon, then quietly shipped. The database works, sign-up works, and the row-level security policies are either missing or wide open. We go through the project table by table and fix the actual holes.
What it is
Supabase is an open-source backend-as-a-service: hosted Postgres with authentication, file storage, a realtime layer and edge functions wrapped around it, reachable from a browser or mobile app through a single client library. It was founded in 2020 by Paul Copplestone and Ant Wilson, and it markets itself as the open source Firebase alternative.
Under the covers it stitches together PostgREST for the auto-generated API, GoTrue for auth, a storage service in front of S3-compatible object storage, a realtime server that streams database changes over websockets, and Deno-based Edge Functions. It has become the default back end for Lovable, Bolt and several other AI app builders, which is why a lot of Supabase projects now exist that no human engineer ever reviewed.
The honest version
No technology is all upside. This is where Supabase earns its place and where it tends to cause trouble.
Sound familiar?
The Supabase problems that bring people to us, and what you have once each one is fixed.
The generated schema has no row-level security policies, or has policies that effectively allow anyone with the public anon key to read and sometimes write every table.
Every table locked down, tested
Policies were written against auth.uid() but the request is reaching the database without a user JWT, so every check fails and the app looks broken.
Auth flow and policies agree
A free-tier project with no traffic for a week gets paused, and clients get connection errors until it is manually resumed or moved to a paid plan.
Right plan for how it is used
Work that should have been a background job is running inside an edge function and hitting the hard execution ceiling.
Moved to the right place
A chatty client, unindexed queries pulling whole tables, or a public bucket serving large files without a CDN in front.
Traffic shaped, costs predictable
Too many channels, a filter that does not match the publication, or a client that reconnects without backoff and gets rate-limited.
Subscriptions trimmed and stable
The hosted daily backup exists, but nothing has ever been restored from it, and point-in-time recovery was never turned on.
Restores tested on a schedule
What we do
Call us for a security review, or hand us the whole project.
How it works
Tell us what's going wrong. We look at it, then tell you plainly what's wrong and what it will cost.
We fix the cause, not the symptom. If something is down, getting it back up comes first.
Backups, monitoring, updates and access, so the same problem can't quietly come back.
You get a plain-language write-up. Or we stay on for hosting, upkeep and new work.
Often found together
We work on the whole stack, so the problem can be wherever it actually is.
Almost always because row-level security is off on the table, or the policies on it allow the anon role to read everything. The public anon key is meant to be shipped to browsers, so the database itself has to enforce who sees what. We go table by table, enable RLS, and write policies that match how the app actually uses the data.
Free-tier projects pause after about a week with no activity. The project is not deleted, but the API and database stop responding until someone signs in to the dashboard and resumes it, or the project is moved to a paid plan that does not pause.
Yes, and it is one of the more common things we are asked to do. We audit the schema, enable row-level security where it was skipped, write policies for the anon and authenticated roles, review the auth flow end to end, and give you a short written summary of what changed and why.
For most small and mid-sized projects the hosted product is the right answer. Self-hosting makes sense when the pricing has grown awkward, when compliance rules require the data to live somewhere specific, or when you need Postgres settings and extensions the hosted plan does not expose.
Edge functions have a hard execution limit of 150 seconds and they cold-start after idle periods. Work that takes longer than that, or that should not block a user request at all, belongs in a background worker outside the edge runtime. We move it there and leave the function doing only what it needs to.
Estimates are free. Describe what the project does and what is going wrong, and you get a price before any work begins.
Free estimate
Tell us what you're running and what's going wrong. A person reads it and comes back with an estimate.
Your message is with a real person, and we'll get back to you soon.
Rather talk now? Call us: