Run exploratory SQL inside isolated sandboxes

Exploratory work benefits from isolation, especially when production systems are shared or constrained. With sandboxes, you can run queries, attach notes, and work with small datasets in a separate environment. This allows you to iterate freely without affecting shared resources.

How it works

Step 1 — Create a sandbox

Create a scratch space called “Staging validation” and switch my session to use it.

Step 2 — Explore tables safely

I’m working inside a sandbox. Show the columns on our product analytics data source, then give me event counts by week for the last 24 weeks from the events table.

Step 3 — Optional sandbox notes

Add sandbox notes: a Notes heading, one bullet with rough row counts from the events table, and one bullet on what I plan to join next.

Step 4 — Optional pilot dataset inside the sandbox

Upload segment.csv as a dataset named high_value_users in this sandbox. What table name do I query afterward?

Step 5 — Create or update a sandbox from the CLI

Start a sandbox called “API exploration” with notes that say “Notebook agent session.”

Step 6 — Close the loop

Tear down or recreate the sandbox when the branch merges or the ticket closes.

Who uses this

• Practitioners prototyping joins before promoting logic to scheduled jobs.
• Operators running agent sessions scoped to a sandbox and sandbox notes.
• Platform teams limiting shared-warehouse user sprawl for experimentation.