The SQL workbench

How the workbench is laid out

On Integrations, a database card’s Open takes you to its workbench. The left side is the schema tree — Tables open by default, Views, Functions and Procedures folded — with a search box and Refresh schema. The middle holds tabs, and a history drawer opens on the right.

1SELECT *2FROM customers3LIMIT 100;

The + menu also searches every table, view, function and procedure by name, so a large schema is one keystroke away. The workbench opens the first table by itself the first time you visit.

Run a query

1. 1
   Open a query tab

   Click + and New query. The editor completes table and column names from the schema it loaded.

2. 2
   Write and run

   Press Run or Ctrl+Enter (⌘+Enter on a Mac). The grid shows the rows, the row count and the time the database took.

3. 3
   Read the warnings

   Above the grid, yellow lines say what could make the query costly. They never stop it from running.

SELECT c.name, COUNT(o.id) AS orders, SUM(o.total) AS revenue
FROM customers c
JOIN orders o ON o.customer_id = c.id
WHERE o.created_at >= now() - interval '30 days'
GROUP BY c.name
ORDER BY revenue DESC
LIMIT 20;

Browse and edit a table

A table tab reads the table with a Filter (column, operator, value), a Show picker for the columns you want, sorting by clicking a header, and a Limit of 50, 100, 200 or 500 rows — 100 by default. Show SQL reveals the exact SELECT it sends.

Edits wait until you apply them

1. 1
   Change what you need

   Double-click a cell to edit it, use Add row (the form marks required fields and lets an auto-increment key be generated), or select rows and delete them. Nothing is written yet: each change is highlighted, and a bar says Editing · 3 changes.

2. 2
   Review the SQL

   Click Apply. A confirmation shows SQL to be executed — every INSERT, UPDATE and DELETE built from your edits.

3. 3
   Apply or cancel

   Apply changes runs the batch and reloads the tab; Cancel on the bar throws every pending edit away.

The tree’s right-click menu creates and changes structure: Edit table opens the table designer (columns, types, primary key, required, default, foreign keys, indexes) with the Generated SQL shown before you apply; Truncate data and Delete table ask first.

Warnings and row limits

Every query is read before it runs and timed after. The answer carries a list of warnings — empty when the query is clean — and the workbench always shows them:

The cap only wraps plain reads (SELECT and WITH). Statements that write — INSERT, UPDATE, DELETE, CREATE, ALTER, DROP — run as written.

History and revert

The Query history button opens every operation run against this connection, newest first, with its status (Succeeded, Failed, Reverted), its time and What was executed. The history is kept on the server per connection — the last 200 entries — so a teammate opening the same workbench sees it too.

• Row edits, inserts, deletes, truncates and dropped tables record the SQL that undoes them. Such an entry is marked Revertible.
• Reverting shows Revert SQL to be executed first; confirming runs it and adds a new entry marked with ↩, linked to the original.
• A revert restores data, not time: rows another process changed in between are not merged.

Limits and access

• PostgreSQL and MySQL only. S3 connections open the file browser instead.
• Running queries needs Edit access to Integrations in the project; see Team access.
• MySQL hides a routine’s body from anyone but its DEFINER. Grant ALTER ROUTINE to the connection’s user to read definitions.