# SQL editor

Run queries against Postgres, MySQL, Turso, Databricks, ClickHouse, and more.

Infrawrench ships an in-app SQL console so you do not need to juggle psql, mysql, and vendor consoles for quick lookups.

![SQL editor with the schema browser expanded and a result table below](https://agent-assets.infrawrench.com/docs-screenshots/features/sql-editor/sql-editor-results.png)

## Where to open it

- **Per-database account** — click the **SQL** button on a Postgres, MySQL, Turso, ClickHouse, or Databricks account.
- **Per-resource** — many managed databases (RDS instances, PlanetScale branches, Neon branches, Cloud SQL, Turso databases) have their own SQL editor on the detail page so you do not have to add a separate account.

## Features

- **Schema browser** down the left-hand side, from live introspection. Filter it with the **Search tables…** box, expand a table to see its columns and types, and click a table name to drop `SELECT * FROM <table> LIMIT 100;` into the buffer and run it. Primary-key columns are marked with a ⚿.
- **Result grid** with primary-key columns badged **PK** and nulls shown as `NULL`.
- **Inline row editing** — where the query selects from a single table and brings back all of its primary-key columns, the pencil on a row turns the non-key cells into inputs and writes an `UPDATE` back.
- **Export CSV** — downloads the current result set.
- **Estimate** — on engines that support a dry run (BigQuery, Databricks), prices the query before you run it; the toolbar then shows the bytes and cost, or "cache hit · free".
- **Run ⌘↵** runs the whole buffer. Tab inserts two spaces rather than moving focus.

The editor itself is a plain text area — there is no autocomplete, and no query history is kept between visits.

## Per-driver notes

- **Postgres** — full SQL, introspection from `information_schema`.
- **MySQL / PlanetScale** — PlanetScale uses Vitess, so cross-shard joins can fail; infrawrench surfaces the raw error.
- **Turso** — libsql dialect. Edge-replicated, so some writes lag on reads until replication catches up.
- **Databricks** — runs against a SQL warehouse; results are fetched via REST, not a long-lived connection.
- **ClickHouse** — HTTP interface. Large result sets stream into the grid.
- **MongoDB** — not SQL; uses a separate collection browser instead (see the [MongoDB plugin](../plugins/mongodb.md)).

## Writing vs reading

There is no read-only mode by default. If you want a query to dry-run, prepend your statement with `EXPLAIN` (Postgres, MySQL, ClickHouse) or whatever your engine supports.