The Data Lab pane

The Data Lab pane turns a file into a quick, honest read on your data. Open it from Tools → Data Lab, the command palette (Open Data Lab), open the data pane, or the Data Lab starter recipe.

Open a dataset

Drop a CSV, TSV, or JSON file onto the pane, click Open file, or Paste raw data. Numbers are detected automatically (id-like values with leading zeros stay text); blank cells read as null.

What you get

• Column profiles — for every column: its type (number / string / mixed), how many values are filled, how many are null, how many are unique, and — for numbers — the min, max, and mean.
• A typed preview — the first 50 rows in a scrollable table, numbers right-aligned, nulls dimmed.

Turn it into work

• Report opens a Writer document seeded with the dataset’s dimensions and a column table, ready for your findings and next steps.
• Analyze hands the dataset — its profile plus a sample — to an Analyst agent. The agent reasons over it and can load the file in its own workspace to compute, chart, and summarize. (An agent CLI must be set up for this.)

Deliberately not a notebook

Data Lab is a viewer, not a code runtime — it never executes code in the app. The trust boundary that already governs agents (their terminal and worktree) is where analysis runs, so nothing new can touch your machine from the pane.