Excel to Markdown
Drop in an .xlsx workbook or a .csv file and get proper GitHub-flavoured Markdown tables - one heading and one table per sheet, pipes and line breaks inside cells escaped so nothing breaks. Everything runs in your browser, and the page shows how many AI tokens the Markdown saves over pasting the same rows as JSON.
Drag & drop a spreadsheet, or click to browse
.xlsx, .csv and tab-separated files, with every sheet in the workbook converted at once. Nothing leaves your browser.
Why convert a spreadsheet to Markdown for AI?
There are only a few ways to get tabular data into a chatbot, and they are not equal. A screenshot forces the model to do OCR before it can even start. JSON is precise but repeats every column name on every row, so a 40-row sheet pays for its headers 40 times over. Raw pasted cells arrive as ragged tab-separated text that models regularly misalign by a column.
A Markdown table names each column exactly once, draws visible boundaries between cells, and uses a syntax every model has seen constantly in training data. That combination is why convert spreadsheet to markdown for ai has become the standard first step before asking a model to summarise, filter or reason over a sheet - and why it is one of the easiest ways to reduce tokens for Claude and every other model without losing any meaning.
What this converter handles
- .xlsx, .csv and .tsv in one tool. Comma, semicolon and tab-separated files are detected automatically, so European exports work without a settings hunt.
- Multi-sheet workbooks. Each sheet gets its own
##heading and its own table, in the original workbook order, and you can switch individual sheets off. - Pipes and line breaks inside cells. A cell containing
Yes | Nowould otherwise end the cell early; it is escaped. A cell with a line break in it would end the row; it becomes a<br>. - Merged cells. Markdown has no merged cells, so a merged range is flattened by repeating its value across every cell it covered - the table stays rectangular and nothing silently disappears.
- Mixed data types. Dates come out as readable
YYYY-MM-DD, percentages keep their percent sign, formulas are replaced by their calculated result, and floating-point noise like0.30000000000000004is cleaned up. - Header rows, or no header row. The first row is treated as the header by default; switch that off and columns are labelled A, B, C instead, because a Markdown table always needs a header line.
Large sheets and token cost
A spreadsheet with thousands of rows will happily produce a multi-megabyte Markdown file that no chatbot can usefully read - most models start losing track of individual rows long before they run out of context. So this tool caps anything over 500 rows by default, tells you the real total, and writes an explicit note under the truncated table saying how many rows were left out. You can raise or remove the limit whenever you actually need the whole thing.
अक्सर पूछे जाने वाले प्रश्न
Does it handle multi-sheet Excel workbooks?
Yes. Every sheet in the workbook is converted, in its original order, each under its own ## heading with its own Markdown table so the model can tell them apart. You can also switch individual sheets off if you only want part of the workbook - useful when a workbook has a data sheet plus a lookup sheet you do not need.
What happens if a cell contains a pipe character or a line break?
Both are handled. A pipe inside a cell is escaped as \| so it renders as a literal pipe instead of ending the cell, and a line break inside a cell becomes an HTML <br> so it does not end the row. This is the single most common reason hand-made Markdown tables come out broken, and it is why the tool escapes rather than strips those characters.
Can it convert CSV as well as Excel?
Yes - .csv and .tsv work exactly like .xlsx, and the delimiter is detected automatically, so comma, semicolon and tab-separated files all work without you choosing anything. A CSV becomes a single Markdown table named after the file.
What about .xls files from Excel 97-2003?
The old .xls format is a binary container that no browser-side reader can open without uploading it to a server, which this tool deliberately never does. If you drop one in, the tool recognises it and explains the fix: open it in Excel, LibreOffice or Google Sheets and save it as .xlsx or .csv.
Is a Markdown table really cheaper than JSON for an AI model?
For tabular data, usually by a wide margin. JSON records repeat every key on every row, so the column names are paid for once per row rather than once per table. A Markdown table writes each column name once in the header and then sends only values. The tool shows the estimated counts for your own data - roughly four characters per token - so you can see the actual difference for your sheet rather than trusting a rule of thumb.
How are merged cells handled?
A Markdown table cannot represent a merged cell, so a merged range is flattened by repeating its value into every cell it covered. That keeps every row the same width, which keeps the table valid, and it means a merged category label still appears next to each of its rows instead of vanishing. The tool tells you how many ranges were flattened.
My sheet has no header row. Will the table still be valid?
Yes. Turn off "First row is the header row" and the tool generates spreadsheet-style A, B, C column labels instead of consuming your first row of data. GitHub-flavoured Markdown requires a header line, so there always has to be one - the choice is only whether it comes from your data or is generated.
Is my spreadsheet uploaded anywhere?
No. The file is parsed in your browser and the Markdown is built there too. Nothing is transmitted, so financial models, client lists and internal reports never leave your machine.