A second opinion for your Postgres, run on your machine.
One command inspects your database read-only, from your own computer. It reads what it finds together and prints what to fix first, why, and how. Your connection string never leaves your machine.
npx pg-checkup@latest postgres://[email protected]/appNode 18+ · Read-only · Nothing uploaded unless you say so
- 3 tables have more than 20% dead rowsVacuum
- fsync is onConfiguration
- Standbys keep upReplication
- 1 inactive replication slot is retaining 6.2 GB of WALReplication
- Remote clients use TLSSecurity
- 2 tables have no primary keyTables
- Sequences are far from their maximumSchema
- Transaction ID age is below the limitVacuum
Works with any PostgreSQL 12 or newer.
- PostgreSQL
- Supabase
- Neon
- Google Cloud SQL
- DigitalOcean
- Railway
- Render
- Amazon RDS
- Aurora
- Self-hosted
Logos are trademarks of their respective owners. pg-checkup is not affiliated with or endorsed by them.
Before you scale it, migrate it, or blame Postgres, check it.
Before you pay for a larger instance
High CPU and slow queries often come from one query, a table full of dead rows, or too many idle connections. The report says which, so you can fix that first and measure again.
Why is my disk filling up, and did my jobs stop?
pg_cron keeps every run forever, pg_net keeps every response, and log and audit tables only grow. A failing scheduled job silently stops cleanups and refreshes. The checkup lists each one. Tables that never clean themselves
Is my database exposed?
Tables and views open to every role, views that skip row-level security, passwordless remote logins, and logins with more access than they need. On any host, not only Supabase.
When it got slow
Findings are read together. A transaction that has stopped vacuum is listed before the dead rows and slow queries it causes, with a note not to add an index yet.
Before and after a migration
Save a baseline of the old database, then compare the new one: missing extensions, policies, grants and functions, sequences that weren't advanced, and tables with no statistics. Migration checks
Three steps
Use your login, or a safer one
Any login works. Optional: a role with pg_monitor reads statistics views and never table rows.
create role checkup_reader login
password 'choose-a-password';
grant pg_monitor to checkup_reader;
-- Supabase: grant usage on schema extensions to checkup_reader;Run it on your machine
The password comes from ~/.pgpass or a hidden prompt. The session is read-only, with a 15-second limit on each query.
npx pg-checkup -h db.example.com \
-U USER -d appRead it, then upload it to share
Findings in the order to fix them, with why and how, in your terminal. pg-checkup upload sends only the statistics file, after asking, and gives you this report as a web page.
FIX NOW
✗ fsync is off
FIX NEXT
! 2 tables have 20%+ dead rowsSample report
Explore a demo database with planted problems, prioritized findings, and the evidence behind them. No signup.
Open the sample reportFind out what's wrong for free. Pay to fix it faster.
Workspace is $12 a month for lasting history, PDF exports, sharing and alerts from scheduled uploads.
Free
- Runs on your machine; nothing uploaded
- Every finding, why it matters and the fix, in order
- Migration baseline and before/after comparison
- Shareable web report with all the evidence
Workspace
- Keep your checkup history together
- Follow changes across retained reports
- PDF and a share link
- Scheduled checkups and alerts
Questions
Do you ever see my connection string?
No. The checkup runs on your machine and connects only to your database. If you choose to upload the statistics file for a shareable report, that file holds statistics, settings and object names, never table rows or credentials, and the command shows you what is in it and asks first.
Is it safe to point at production?
The session is forced read-only, each query has a 15-second timeout, and no table rows are read. Any login works; a read-only login (pg_monitor) is optional extra safety, and a login that can read tables only gets a warning. pg-checkup sql prints every query it runs.
Which databases work?
PostgreSQL 12 or newer, anywhere you can reach from your machine: a laptop, a bastion host, CI. Node 18 or newer runs it.
Is this monitoring?
Not live monitoring. It checks when you run it, or on a schedule you set. Run it after a fix or before a deploy, and compare migration baselines or traffic windows with the baseline, verify, observe and compare commands.
My provider already has a performance dashboard. Why use this?
Your provider's tools show how its service is running. pg-checkup reads what Postgres itself records and reads the findings together: when a long transaction stops vacuum and that slows a query, the report says to end the transaction first, not to add an index. It works the same on every host, so it also fits before and after a move between providers.
What does Workspace add?
Findings, SQL recommendations and rechecking are free. Workspace keeps your report history together and adds PDF exports, sharing, scheduled uploads and email alerts. Workspace is $12 a month.