Skip to main content

pocketPG

Agentless PostgreSQL diagnostics that tell you what's wrong, prove it with evidence, and hand you the SQL to fix it.

pocketPG connects directly to a PostgreSQL instance the way psql does — no collector, agent, or sidecar to install — reads the standard catalog and statistics views, and turns them into ranked, explained, and actionable findings.

Why pocketPG is different​

Most Postgres monitoring is either a wall of graphs you have to interpret yourself, or a black box that fires alerts you can't verify. pocketPG is built on four opinions.

1. Agentless by design​

There is nothing to deploy next to your database. pocketPG connects directly, or through an SSH tunnel for private instances — no collector process, no daemon, no package to keep patched. Onboarding a database is a connection string, not a rollout.

2. Metrics you can trust​

Rates and ratios are measured over a recent window, not averaged over the server's entire uptime since the last statistics reset. A database that has been up for months no longer reports a comfortable 99.9% cache-hit ratio while the last few minutes are actually thrashing. When pocketPG cannot yet measure a real window, it says so and refuses to raise a false alarm — a cumulative number never triggers a CRITICAL.

3. Every finding comes with a fix​

A finding is not just "connection saturation is high." It carries the exact statement to address it — an ALTER SYSTEM SET …, a VACUUM (FREEZE …), a REINDEX … CONCURRENTLY — together with whether a restart is required and the caveats that matter. Fixes are always copy-to-run: pocketPG surfaces the SQL in the app, the daily digest, and webhook alerts, but never executes a change on your database itself.

4. Honest about certainty​

Some signals are provable from SQL; others are heuristics. pocketPG labels the difference on purpose. Recommendations carry their supporting evidence and a rollback path, so you decide — you are not asked to obey a score you can't inspect.

What it's for​

  • Health & incident triage — findings ranked by severity across checkpoints, WAL, wraparound, connections, vacuum, replication, and risky configuration, each with a fix.
  • Query analysis — from pg_stat_statements, with plan inspection and index guidance.
  • Configuration tuning — workload-aware recommendations that ship with evidence and rollback SQL.
  • Extension monitoring — dedicated checks for pgvector, PostGIS, and logical replication.

It is deliberately not a full metrics platform, a failover orchestrator, or a compliance suite.

Start here​

  1. Connect your database — directly or via SSH tunnel
  2. Overview tab — the whole instance at a glance
  3. Health dashboard — ranked findings, each with a fix
  4. How recommendations work — evidence and rollback, not black-box advice

Everything else is in the sidebar, grouped by Getting started, Monitoring, Query analysis, Tuning, Vacuum & bloat, Extensions, and Operations.

What pocketPG reads​

Standard catalog and statistics views only — pg_stat_database, pg_stat_activity, pg_stat_statements (when installed), pg_stat_user_tables and pg_stat_user_indexes, pg_stat_wal, pg_settings, and selected pg_catalog helpers. Every page states which views it needs and what happens when a view or privilege is unavailable — rather than failing silently.