How it works

AskCSV turns plain-English questions about a delimited file into numbers, tables, and charts that can be checked. Guest use needs no account, and the generated SQL is visible for every chart.

How do I start analyzing a file?

Open the file input, choose a CSV or Excel file, or paste delimited rows. Use a header row with meaningful names such as region, revenue, and cost — clearer headers produce better queries.

The file is opened in this browser and kept in this site's storage for your account, or for a guest session if you have not signed in. It is not uploaded. There is no setup step, no model to download and no graphics requirement, so the first answer does not wait on an install.

Does the AI read every row to calculate an answer?

No. The model never gets the file; what it sees is capped and kept to a minimum. The workflow separates understanding the question from calculating the answer:

1. Understand the columns. The AI provider receives your question, the file's name and row count, the names and types of the columns you chose to share, up to 10 sample rows, and the earlier messages in the chat, so it can tell a date column from a currency column.
2. Generate a query. It proposes a read-only SELECT for your question. AskCSV checks the SQL against its allowed-query and column rules before execution.
3. Calculate with DuckDB. The browser runs the query against the local file, over all matching rows — not a sample. Totals come from query execution, not from guessing.
4. Compute and explain. The app computes summary facts in code from the query result, and the AI describes them. Charts, tables, and KPI figures are built from the same result.

To explain, the AI receives the query result — up to its first 100 rows (16 KB), from the columns you share — with short facts computed from the whole result, such as its total and its highest and lowest rows. It never receives the file itself, but a result can hold real rows: ask for the five largest orders and those five rows go.

What does that look like for a sales question?

Suppose a fictional export contains region, gross_sales, and refunds. Ask: "What is total gross_sales by region? Sort from highest to lowest."

A representative calculation would be:

SELECT region, SUM(gross_sales) AS total_gross_sales
FROM sales
GROUP BY region
ORDER BY total_gross_sales DESC;

Here, sales is an illustrative table name, not something you type into AskCSV. The model uses the schema and sample rows to interpret the fields; DuckDB performs the sum across the whole dataset. This is an explanation of the method, not a recorded AI output. The same SQL is shown alongside the chart in the app, so the version that actually ran is the one you can read.

Notice what the question does not specify: subtracting refunds. If you want net sales, ask explicitly for gross_sales minus refunds, summed by region. A valid SQL query can still answer the wrong business question. Check the measure, filters, grouping, and returned numbers before acting on the explanation. For a hand-checkable exercise, use the six-row CSV summary example.

Which file formats are supported, and what does it cost?

AskCSV accepts CSV, TSV, delimited TXT, and Excel .xlsx, including cells pasted from a spreadsheet. For a workbook it reads every sheet's values as its own table — not formatting or live formulas. Older .xls files need to be saved as .xlsx or CSV first. Report titles above the table, blank rows, and a trailing total row are cleaned up when the file opens, and the app tells you what it changed. AskCSV analyzes opened files only — there is no direct database connection.

How large a file can be handled depends on the browser and device rather than on a plan quota. Only AI answers use credits, charged for the tokens the AI reads and writes. See pricing.

What data is sent where?

Opening a file does not send it anywhere: it stays in this browser. The destinations worth naming are for the question, and AskCSV keeps what goes to the AI to a minimum: through AskCSV's server, the AI provider (OpenAI) receives only your question, the file's name and row count, the columns you chose to share with up to 10 sample rows, the earlier messages in the chat, and up to 100 rows of each query result plus summary figures. The server keeps a hash of the request context, token counts and the charge — not the question, sample, result or answer text. Google handles sign-in, Stripe handles payments, and PostHog receives cookieless aggregate analytics. AskCSV shows no ads. The web app is hosted on Vercel, and the database runs on Oracle Cloud Infrastructure.

Your files and chats stay in this browser until you delete them or clear the site's data. Guest accounts and trial credits expire 7 days after they are created. You can delete any file or chat at any time. The data-handling walkthrough and the privacy policy cover this in more detail.

Frequently asked questions

Will it work in any browser?

Yes. Files open and queries run in your browser, so any modern browser works, including on phones and tablets. There is no model to download and no graphics requirement. An internet connection is needed for a new AI answer.

Are answers guaranteed to be correct?

No. SQL validation limits what a query can do; it does not establish that the AI understood your intent. Verify consequential results against a known calculation, and use the displayed SQL to check what ran.

Does AskCSV make my workflow compliant?

No. Browser-side processing with a disclosed sub-processor, read-only queries, and clear retention are useful facts for a review, but they do not certify GDPR, HIPAA, or any other framework. Your organization must evaluate its own obligations. The data minimisation checklist is a starting point.

Technical references

• DuckDB documents the SQL engine AskCSV queries local files with.
• DuckDB aggregate functions explains COUNT, SUM, and AVG, which most answers rely on.

Try the workflow with a CSV or pasted sample.