OUR METHOD / NO MAGIC NUMBERS

How QuerySense reasons.

A useful optimization tool must show what it knows and what it cannot know. QuerySense builds a trail from SQL structure to PostgreSQL plan evidence, then asks you to validate a controlled change.

01 / PARSE

Inspect the statement

One PostgreSQL SELECT at a time. The parser rejects malformed or unsupported SQL. Structural patterns become cautious findings; they are not runtime measurements.

02 / OBSERVE

Read the plan

Paste EXPLAIN JSON from your own database client. Estimated plans show planner choices and costs; ANALYZE plans can add actual rows, timing, and buffers. QuerySense never connects to your database.

03 / TEST

Compare experiments

Paste baseline and candidate plans to compare observed time when both are measured, or planner cost when they are not. Repeat trials and confirm result equality before deploying. Optional AI adds hypotheses tied to findings.

Privacy and limits. Analysis requests are processed without server-side storage. Saved inputs use opt-in browser storage. AI review sends supplied SQL, schema context, and baseline plan summary to the configured provider only when enabled. A pasted ANALYZE plan contains measurements from your database; QuerySense does not measure the query itself. Planner cost is not elapsed time.

Based on the PostgreSQL guides to using EXPLAIN and planner statistics.

Try the workbench