To add one thing to check before support replies: "Failed to fetch" with no status code means the browser never got a response, so the request is often blocked on your side rather than refused by Supabase. - Open DevTools > Network, reload the dashboard and click the red api.supabase.com request. If it shows "(blocked:other)", "ERR_BLOCKED_BY_CLIENT" or a DNS error, an ad blocker, privacy extension, VPN, corporate proxy or DNS filter (Pi-hole, NextDNS, router "security" features) is blocking it. - Try one browser with no extensions on mobile data (not Wi-Fi), so you're on a different network and DNS. If it works there, the cause is your network or extensions. - If it fails everywhere, include the exact Network error in your email to support@supabase.com. It points them straight to the account side.
Since SQL and auth work and only /rest/v1 fails, PostgREST is most likely failing on startup while it loads your schema. That's usually caused by something in the database, not the platform. Worth checking while you wait on the ticket: 1. Recent migrations. A broken view or function in an exposed schema (often a view pointing at a dropped column, or a function with an invalid return type) can stop the schema cache from loading. Run `notify pgrst, 'reload schema';` in the SQL editor, then check Logs > PostgREST for the actual error message. 2. Settings > API > Exposed schemas. If a listed schema was dropped or renamed, PostgREST can't start. 3. The `authenticator` role. If a migration changed its password, `statement_timeout` or grants, PostgREST can't connect even though Postgres is fine. 4. Connections. Check whether you're at your connection limit (`select count(*) from pg_stat_activity;`). pg_cron jobs plus an app pool can starve PostgREST. The PostgREST log line will usually name the failing object. If you paste it here with keys removed, people can point to the fix.
Here's a quick way to find the actual error, which is the first step: 1. Open the page, press F12 (or right-click > Inspect), and look at the **Console** and **Network** tabs. Reload the page. Click any request in red and note the status code and response. 2. What the common results mean: - **Blank page, nothing in Network:** usually a frontend build problem, such as a wrong base path or missing env vars at build time. Check that `VITE_SUPABASE_URL` / `NEXT_PUBLIC_SUPABASE_URL` and the anon or publishable key are set on your host (Vercel, Netlify and so on), then redeploy. Env vars are baked in at build time. - **401 "Invalid API key":** the key in your hosted env is wrong or was rotated, or legacy keys were disabled in Settings > API Keys. Use the current publishable key. - **42501 "permission denied for table":** the table was created after Supabase stopped auto-granting new tables. Run `grant select on public.<table> to anon, authenticated;` and keep RLS on. - **200 with `[]`:** RLS is on but no policy allows the read. Add a `select` policy. - **CORS error:** you're calling something other than the Supabase URL, or an edge function isn't returning CORS headers. - **Project paused:** free projects pause after inactivity. Restore it from the dashboard. 3. In the Supabase dashboard, check Logs > API for the same request. If you post the status code and error message (blank out keys), people here can point to the exact fix.
I haven't done a full migration myself, but I see the same few things go wrong in apps that have been moved onto their own Supabase project: 1. **Auth and passwords.** If you can't export password hashes from the old platform, users have to reset their passwords. Plan a "set a new password" email instead of assuming logins will carry over. Also add every production and preview URL under Auth > URL Configuration (Site URL plus Redirect URLs). A missing entry here causes most of the "login works locally, loops in production" reports. 2. **Table grants.** Since 30 Oct 2026, new tables in `public` aren't granted to `anon`/`authenticated` automatically. If you recreate tables with a migration, add `grant select, insert, update, delete on public.<table> to authenticated;` (and `anon` only where needed) on top of your RLS policies. Otherwise you'll get `42501 permission denied`. 3. **RLS.** Platforms that managed the backend for you sometimes had RLS off or used wide-open policies. Turn RLS on for every table and write the policies before you go live. Empty arrays with no error almost always mean RLS is on but no policy matches. 4. **Secrets.** Move any service-role or secret key out of frontend code and into server or edge-function env vars. Then rotate it, because the old one has been in a public bundle. 5. **Storage.** Re-create buckets and their `storage.objects` policies. Files don't move with the database dump. From day one, I'd keep schema, grants and policies in migrations (`supabase/migrations`) so the whole backend can be rebuilt from git. More detail on the 30 Oct change: https://mendlark.com/guides/supabase-october-30-change-what-breaks.html