Paste a table from Claude — a markdown table in the chat, a CSV artifact, or an ASCII grid — and copy it into Excel as real cells, or download a .xlsx with numbers, percentages and currency typed, a bold frozen header and one sheet per table.
Select the table in Claude, copy, and paste it into the box above. It is detected as a grid; click Copy for Sheets / Excel and paste into Excel for separate columns with numbers recognised, or click Download .xlsx for a typed workbook. CSV from a Claude artifact and ASCII tables with +---+ borders are read the same way. Everything runs in your browser.
Use the copy button under the answer or select the text. Claude's copy gives markdown text; either form is read correctly here.
The table is detected as you paste and shown as a grid. Switch tables with the tabs if the answer had more than one.
Click Copy for Sheets / Excel and paste into Excel, or Download .xlsx for a typed file.
Claude produces tables in three different forms, and each pastes into Excel badly in its own way. In the chat, a table is markdown: rows of text with cells divided by | and a line of dashes under the header. Copy it and Excel puts each row into one cell, pipes included. When you ask for data you can import, Claude often answers with a CSV block or a CSV artifact — comma-separated lines with quoted fields — which Excel will open from a file but, pasted, lands in one column. And when Claude is explaining a terminal output or a plain-text report, it draws an ASCII table with +---+---+ borders, which no spreadsheet reads at all.
This page detects all three. Markdown pipes are split on the unescaped | characters and the dashes row identifies the header. CSV is parsed by the RFC 4180 rules, so a quoted field containing a comma, a line break or a doubled quote stays one cell. ASCII borders are treated as separators and the border under the first row marks the header. Whatever the shape, you get the same grid, the same preview and the same three routes into Excel.
Claude likes a helpful table. It writes 12 kg rather than 12, ~$4,200 rather than 4200, N/A for a missing value and a footnote marker after a figure it is unsure of. For a reader that is considerate; for a spreadsheet it means the column is text and the total is zero. The number detection here reads the common forms a spreadsheet user expects — thousands separators, a leading $, € or £, a trailing %, parentheses or a minus for negatives — and types those cells as numbers in the .xlsx. A value with a unit attached stays text, which is the honest choice; ask Claude for a separate unit column, or strip the units in the preview, and the column adds up.
Claude also bolds the first column of a comparison table and occasionally the header cells, with **asterisks**. The markers are removed from the cells on paste, and the header row is written bold in the .xlsx regardless. Hidden characters — the narrow space Claude puts between a number and its unit, non-breaking spaces from a copied web source — are stripped before the number detection runs, which is the usual reason a pasted figure refuses to sum.
A Claude analysis often contains more than one table: a summary, then a breakdown by month or by region, then a sensitivity table. Paste the whole answer. Every table is found separately and listed in the tabs above the preview; prose between them is ignored. Copy for Sheets / Excel and the CSV download work on the table you have selected, and Download .xlsx writes all of them, one per sheet, in the order Claude wrote them. That is usually the fastest way to turn a Claude analysis into a workbook you can build on.
Two prompts make the conversion cleaner. "Give me the table as CSV with a header row, numbers only in numeric columns, units in the column name" produces data that types perfectly. And "one table per section, no notes inside cells" keeps footnotes out of the figures. Claude follows both reliably, and the result pastes into Excel with every column numeric and nothing to fix.
Yes. Copy the artifact's contents with the panel's copy button and paste here; comma- and semicolon-separated text is parsed by the CSV rules, including quoted fields with commas or line breaks inside them. Then copy for Excel or download the .xlsx.
Because a cell like 12 kg is not a number, and silently dropping the unit would be wrong. The .xlsx keeps it as text. Ask Claude to put units in the column header instead, or edit the cells in the preview, and the column becomes numeric.
Yes. 8.5% is stored as 0.085 and formatted as a percentage; $4,200 is stored as 4200 with a dollar format, so the column sums and still displays as currency. Codes with leading zeros and very long digit strings stay text on purpose.
They are detected. The +---+ borders are treated as separators, the border under the first row marks it as the header, and the cells are read like any other table.
No. The conversion runs in your browser with client-side JavaScript. The Claude answer is never sent to a server, stored or logged.