Turn a CSV into a Markdown table

A Markdown table is for reading, not for storage. 50,411 rows of CO2 data in a README is unreadable and GitHub stops rendering long files anyway. So the Markdown export writes the first 1,000 rows of your current view, and the point of this page is getting the right 1,000.

If someone needs the whole dataset, share a link next to the table: curl -T data.csv https://rows.page returns one, and it opens as a filterable table for anyone.

Export writes what you see. country:Netherlands year:>=2010 keeps 15 rows, hiding 70 of the 79 columns keeps the table narrow, and clicking a header sets the order. Shift-click sorts by a second column.

For a summary table, run SQL in the same bar: SELECT country, round(max(co2)) AS peak_co2 FROM data GROUP BY 1 ORDER BY 2 DESC LIMIT 20 gives a 20-row ranking that pastes straight into a pull request description.

Pipes inside a value become \|, backslashes are doubled, line breaks are replaced with a space, and nulls become empty cells. The header row uses your column names and the second line is the --- separator GitHub, GitLab, Obsidian and most Markdown renderers expect.

Numbers and dates are written as text exactly as the table shows them. Nested JSON columns are written as JSON text in the cell.

A Markdown table is the most token-efficient way to hand a model a few hundred rows while keeping the column names attached to every value. Filter to the rows the question is about, export Markdown, paste. For bigger inputs, export CSV or JSONL instead and attach the file.

Agents can skip the paste: the MCP server at rows.page/mcp pushes a file and reads it back with filters, so the model only pulls the rows it needs.

The Export menu also writes CSV, Parquet with zstd compression, a JSON array and JSONL, all of the current view and without the 1,000-row cap (SQL results stop at 1,000,000 rows). The CO2 sample is 13.7 MB as CSV and 4.3 MB as Parquet.

Everything runs inside your browser tab, so the 1.5 GB limit is the memory of the tab, and the file never leaves your machine unless you press Share.