0.17.0

Reading data through an AI client has always meant reading its description of your data. You asked for last month's invoices and got a paragraph, or a wall of JSON, and anything you wanted to check yourself meant asking again and waiting again.

Results now arrive as a real table, rendered inside the conversation.

Sort it, filter it, then ask about it

Click a column to sort. Type in the filter box to narrow hundreds of rows to the three you care about. None of that goes near the model, so none of it costs you anything or makes you wait.

What you narrow to is shared with your assistant automatically. Filter to two overdue invoices, then just ask why they are overdue. There is no button to press and nothing to copy back into the chat.

Charts where they help

Two columns and a number turn into a bar chart. A single value renders as a single value instead of a one cell table. Your assistant can also pick the view when it knows what you are looking for, and you can always override it.

You can see which apps the data came from

Every result shows the connected apps behind it. That matters most when there is more than one, because a single question can now span them. Ask about jobs and the invoices raised against them and you get one table drawn from both, with both apps named on it. No individual app can show you that, because no individual app has the other's data.

Where it works

In Claude and in ChatGPT, on web, desktop and mobile. Clients that cannot render a table are unaffected and keep receiving exactly what they received before.