Docs

pgblame is two pieces: an agent that runs next to your Postgres, and a hosted dashboard. To get going:

1. Sign up and create a project — you get an agent token.
2. Run the agent next to your Postgres. Pick your provider:
   • Supabase
   • Neon
   • AWS RDS / Google Cloud SQL / Azure
   • Self-hosted (VM, Docker, Kubernetes)
3. Point your deploys at us: Vercel, Railway, or a curl one-liner from any pipeline.

What it does, in 30 seconds

Every minute, the agent reads pg_stat_statements from your Postgres, computes per-query deltas vs the previous tick, hashes the query text, and ships a small JSON payload to pgblame.com/v1/snapshots. When you deploy, your CI hits pgblame.com/v1/deploy with the commit. The dashboard joins them so you can see which queries got slower right after the deploy.

It never sees your application data. Read the setup doc for the literal SQL the agent runs.