Editor’s note — updated 1 October 2026

This post announced the first release. A few details have changed since then (see the changelog):

Administrative statements: GRANT, REVOKE and user management are still blocked outright, but CREATE EXTENSION, DO and SET GLOBAL / SET PERSIST are now refused only for untrusted procedural languages and dangerous variables. See SQL guards.
Self-approval: a system administrator can approve their own request; the override is recorded in the audit log. See Approval workflow.
Roles: Admin is a system-wide role; DBA, Developer and Auditor are assigned per team.
First account: there is no web sign-up. Create the first administrator with QUERYPROXY_ADMIN_EMAIL / QUERYPROXY_ADMIN_PASSWORD on first boot or php artisan queryproxy:create-admin, and install with a single docker run. See Quick start.

Every engineering team eventually has this conversation: a developer needs production data to debug something real, and the only ways in are handing over credentials, opening a tunnel, or a DBA running queries on other people’s behalf all day. All three are bad — the first is how incidents happen, the last two don’t scale.

QueryProxy is our answer: a self-hosted portal that sits between your team and your databases, replacing shared passwords with a workflow.

How it works

A developer writes SQL in a guarded editor, against a connection a DBA granted them. Credentials stay in an AES-256-encrypted vault.
Before the request is even submitted, the SQL is parsed and inspected: UPDATE/DELETE without a WHERE is rejected, SELECT without a LIMIT gets one injected (default 1000, hard cap 10000), and administrative statements are blocked outright.
A DBA approves or rejects — from the web queue, or straight from Slack with interactive buttons. Every chat callback is HMAC-SHA256-verified with replay protection, and nobody can approve their own request (a system administrator can override this, and the override is flagged in the audit log).
A queue worker executes the query asynchronously, streaming results row-by-row through your masking rules — matching values are masked before they touch disk — and writing every step to an append-only audit log.

What ships today

Teams with Admin / DBA / Developer / Auditor roles and hard isolation
PostgreSQL, MySQL, MariaDB, SQLite and SQL Server targets (SQL Server needs a custom image: the published one ships without the driver)
SQL guards enforced by parsing, not pattern-matching
Slack interactive approvals and a replay-protected Teams HMAC action endpoint
Column-pattern and content-regex data masking (full / partial / hash)
Optional TOTP two-factor authentication, enforceable per role
Streaming NDJSON results, CSV export, automatic retention pruning
An append-only, filterable, exportable audit log (the application never edits or deletes entries)

Try it in two minutes

docker run -d --name queryproxy \
  -p 7432:7432 \
  -v queryproxy-data:/var/www/html/storage/app \
  -e [email protected] \
  -e QUERYPROXY_ADMIN_PASSWORD='choose-a-strong-password' \
  queryproxy/queryproxy

Open http://localhost:7432 and log in with that address. The admin variables create the first account on first boot; without them, create it with docker exec -it queryproxy php artisan queryproxy:create-admin. To explore with a demo team instead, set QUERYPROXY_SEED_DEMO=true together with QUERYPROXY_SEED_DEMO_FORCE=true (the image runs with APP_ENV=production, which refuses the seeder otherwise) — for demo instances only, not for production. Its one-time password is printed in the container logs. The documentation covers everything from quick start to Slack setup and the exact guard rules.

QueryProxy is free and open source under AGPL-3.0. Star it, fork it, run it tonight — and if you find something sharp, the issue tracker and security policy are waiting.