Supabase says too many connections. Here is why, and the fix
Postgres caps how many clients can connect at once with max_connections, and reserves a handful of those slots for superuser roles. On hosted Supabase the postgres role you connect as is not a superuser, so when ordinary connections run out, your app is refused too — even though a few slots are technically still free.
Last checked November 18, 2026.
What the error actually means
The exact message is usually one of two things. "FATAL: remaining connection slots are reserved for roles with the SUPERUSER attribute" means every ordinary connection slot is in use and only the superuser-reserved ones are left, which your app's role cannot use. "FATAL: sorry, too many clients already" comes from the pooler (PgBouncer/Supavisor) itself, when the number of client connections exceeds its own max_client_conn ceiling — a different, usually much higher limit than the database's own max_connections.
Both are capacity errors, not data-loss errors: nothing is corrupted, and connections that do get in keep working. New connection attempts are refused until something frees a slot.
The real limits, by compute size
Supabase documents a fixed database connection cap and a separate, larger pooler cap for every compute add-on.
| Compute size | Direct database max connections | Pooler max clients |
|---|---|---|
| Nano (free) | 60 | 200 |
| Micro | 60 | 200 |
| Small | 90 | 400 |
| Medium | 120 | 600 |
| Large | 160 | 800 |
| XL | 240 | 1,000 |
| 2XL | 380 | 1,500 |
| 4XL | 480 | 3,000 |
| 8XL | 490 | 6,000 |
| 12XL | 500 | 9,000 |
| 16XL | 500 | 12,000 |
Sources: Supabase: Compute and disk
The fix: use the pooler, not the direct connection
- Switch your app's connection string to the pooler endpoint (the *.pooler.supabase.com host, Session mode on port 5432 or Transaction mode on port 6543) from Database settings → Connection pooling in the dashboard. The pooler's client-connection ceiling is separate from, and much larger than, the database's own max_connections, so it absorbs far more concurrent app instances.
- In a serverless function or edge function, keep one connection (or a pool of one) per instance rather than opening a fresh one per invocation; a burst of cold-starting functions is the single most common way to exhaust the direct-connection cap.
- Reuse your database client across requests in a long-running server instead of opening a new client per request.
- If you are still over the pooler's own limit, find the offending connections with pg_stat_activity, close anything idle, and only then consider a larger compute tier.
- Transaction mode (port 6543) does not keep session state between statements (no prepared statements, no SET persisted across calls); use Session mode (port 5432) if your ORM needs that.
Sources: Supabase: Connection pooling and limits, Supabase: Connect to your database
How Scribase avoids the same wall
Every Scribase project runs behind a pooler by default for application traffic, and a project is $25 a month on Pro with no separate connection add-on to buy mid-incident. Direct Postgres connections from the internet (ports 5432/6543) are not open yet; migrations, the console SQL editor and the HTTP serverless driver work today and do not compete with your app's own connection budget the way a direct psql session would.
Sources: Scribase pricing
Questions people ask
What does "remaining connection slots are reserved for roles with the SUPERUSER attribute" mean?
Postgres reserves a handful of connection slots so an administrator can always get in, even when the server is at its max_connections limit. On hosted Supabase, the role your app connects as is not a superuser, so once ordinary slots run out, your app is refused even though those reserved slots are technically free.
Is "sorry, too many clients already" the same error?
No — that one comes from the pooler itself (PgBouncer/Supavisor) when the number of client connections exceeds its own max_client_conn setting, which is a separate, usually much larger number than the database's direct max_connections.
What is the fastest fix?
Point your app at the pooler connection string (Database settings → Connection pooling) instead of the direct one. The pooler multiplexes many client connections onto a smaller number of database backends, so it absorbs far more concurrent app instances than the direct connection cap allows.
Does upgrading my compute size raise the limit?
Yes. Supabase's documented direct connection caps rise with compute size, from 60 on Nano/Micro up to 500 on 12XL/16XL, with the pooler cap rising in step from 200 up to 12,000.