Spreadsheets you can question
Drop a CSV or JSON into the chat and it becomes a queryable table. Ask in plain language; get a number, a table or a chart.
The bank export, the sales sheet, the workout log — the data you actually have lives in files nobody opens twice. Give the file to the assistant and it becomes a table you can ask about for years.
- expenses-2026.csv
Here's my bank export
Imported 312 rows into the table
expenses— 4 columns, dates from January to August.How much did I spend on restaurants in August?
€412.30 across 17 visits. The top three:
Restaurants, August 2026
merchant visits total Trattoria Nonna 5 €148.00 Bao House 4 €96.50 Café Lumen 3 €61.20 14 more rows
Ask, don’t filter
Questions become queries behind the scenes. The answer comes back as the right shape: a single number, a small table, or a chart when you ask for one.
Chart my spending per merchant this quarter, top five
Top five merchants, July to September.
Spend per merchant, Q3 2026
- Rewe
- €612
- Deutsche Bahn
- €388
- Amazon
- €297
- Trattoria Nonna
- €214
- Spotify
- €36
How it works
There is no “create table” step. Posting data to a table name creates it, infers the columns and types, and stores the rows in a form built for fast analytics. Later posts to the same name append and extend the columns. Queries run on an analytical engine; every project’s tables are kept apart from every other project’s, so a query can never see another project’s data. Changes the assistant makes to rows are previewed first and can be undone.
In the app
The Data tab lists your tables with columns and row counts, lets you add rows and preview them. Queries run from the chat; a result can be opened in full or fed into an automation.
Saved charts Soon
A chart the assistant drew in reply to a question, pinned to the project as a card you can come back to.
Questions
What formats can I upload?
CSV and JSON, as a file in the web app or pasted into the chat. The columns are inferred from the data; later uploads to the same table extend it if new columns appear.
How big can a table be?
Tables are stored in a form built for analytics and queried by an analytical engine, so millions of rows are fine. Your plan's storage cap is the limit.
Can I write the query myself?
Yes — type it into the chat and it runs as is, or use it inside an automation. When you ask in plain language the assistant writes the query for you, and you can ask to see it.
Related
Features
Reports the AI writes, your browser runs
Describe a recurring report or check. The AI writes a small program that runs on your data inside your browser — nothing leaves it.
Use cases
Personal finance without a spreadsheet habit
Export the bank statement once a month, ask questions the rest of the time. A weekly spend report writes itself and runs in your browser.
Guides
How to ask questions about a spreadsheet in plain language
Once your data is a table, questions become answers: numbers, small tables, charts. How to phrase them and what to expect back.