pgblame — which deploy slowed your Postgres?
Log inStart free
◆BACKED BY PG_STAT_STATEMENTSWhich deploy<br>slowed your<br>Postgres?<br>A Docker agent snapshots pg_stat_statements every minute and ties query regressions to your Vercel and Railway deploys. When something gets slow, you get the commit that did it.<br>▷ Start free5-min setup
since last deploy · 14 minutes agopgblame ▸ live<br>SELECT id FROM messages WHERE thread_id = $1 ORDER BY created_at DESC LIMIT 5012ms280ms↑23×SELECT count(*) FROM users WHERE active = true3ms3ms—INSERT INTO events (kind, payload) VALUES ($1, $2)2ms2ms—└ vercel · main · 0bf42c9 · "index on messages(thread_id, created_at)"<br>$ pgblame agent ▸ sampling pg_stat_statements every 60s
Works with<br>Supabase<br>Neon<br>Vercel<br>Railway<br>GitHub Actions
03 · How it works<br>Three steps.<br>Five minutes.
No agents to deploy. No code changes. No language SDK. Just a Docker container next to your Postgres and a webhook from your CI.
◇ Step 01<br>Run the agent<br>$ docker run -d \<br>-e PGBLAME_DATABASE_URL=… \<br>-e PGBLAME_TOKEN=pgb_… \<br>ghcr.io/liberzon/\<br>pgblame-agent:latest<br>25 MB image · static Go binary · read-only role
◇ Step 02<br>Wire your deploy webhook<br>Vercel and Railway are two clicks. GitHub Actions or any other CI is a one-line curl.
Pre-built integrations · or POST to /v1/deploy
◇ Step 03<br>Open the dashboard<br>Every query that got slower (or faster) since the deploy, with the commit and author. Email and Slack alerts on regressions.
30-day history on Pro · 7 days on free
04 · Honest comparison<br>Use the right<br>tool. Maybe not us.
Lantern is great if you're on Rails. pganalyze is great if you have a budget. We built pgblame for the gap between them — language-agnostic, deploy-aware, $19/mo.
ToolPriceDeploy correlationHistoryNon-RailspgHerofree—current state only✓Lantern$9/mo✓30 days—pganalyze$149+/mo✓long✓Supabase Query Performancefree w/ Supabase—current state only✓pgblame$19/mo (free tier)✓30 days✓
05 · Pricing<br>Free to start.<br>$19 to keep going.
The free tier is a real product, not a trial. Pro adds projects, history, and alerts when you need them.
◇ Free<br>$0<br>/forever
✓1 project<br>✓7-day query history<br>✓Now + Since-last-deploy views<br>✓No card required<br>Start free
◇ Prorecommended<br>$19<br>/per month
✓5 projects<br>✓30-day query history<br>✓Email alerts on regressions<br>✓Slack and Discord webhooks<br>✓Priority support<br>Start free, upgrade later
06 · FAQ<br>Questions<br>we keep getting.
Do you ever see my data?No. The agent runs in your environment and reads only aggregate stats from pg_stat_statements — never your application rows. The connection string and credentials never leave your machine. The agent is MIT-licensed at github.com/liberzon/pgblame-agent — you can read the literal SQL it runs.<br>Will the agent slow my database down?Vanishingly. One SELECT every 60 seconds against pg_stat_statements (a system catalog view). The session has a 5-second statement timeout and runs as a non-superuser, read-only role. If the query takes more than 2 seconds three times in a row, the agent backs off automatically.<br>Does it work with Supabase / Neon / RDS / self-hosted?Yes. The agent only needs a Postgres connection string and a role with pg_read_all_stats. Supabase exposes it directly; Neon, RDS, and self-hosted Postgres are all supported.<br>How is this different from pgHero / Lantern / pganalyze?pgHero is current-state only — no history, no deploy correlation. Lantern does deploy correlation but is a Rails-only gem. pganalyze does the most but starts at $149/mo. pgblame is the same shape as Lantern, language-agnostic, at $19/mo.<br>Can I self-host the dashboard?The agent is MIT-licensed and the schema is open. The hosted dashboard isn't officially supported as a self-host target at MVP — fork at your own risk. Most users find $19/mo cheaper than running it themselves.