Markdown Table Extractor
Paste a whole README or doc — every GitHub-flavored pipe table in it is found and exported as CSV, JSON or JSON Lines. Tables inside code fences are ignored. Nothing is uploaded; it all runs in your browser.
Pull every table out of a Markdown document
Paste a whole README, changelog or spec — not just a lone table. Every GitHub-flavored pipe table in it is found, in document order, and exported as CSV, JSON, JSON Lines, or a list of what the document contains. Everything runs locally in your browser; the text is never uploaded.
A table is recognised the way GitHub recognises one: a header row containing
pipes, followed directly by a separator row (| --- |, |:---|, |---:|,
|:--:|) with the same number of cells. Outer pipes are optional, and pipe
lines inside ``` or ~~~ code fences are skipped — so the SQL snippet in
your README is never mistaken for data.
Worked example
Input:
# Release 2.1
## Downloads
| file | size |
| --- | ---: |
| app-linux.tar.gz | 12 MB |
| app-macos.zip | 14 MB |
## Plans
| plan | seats |
|:---|---:|
| Free | 3 |
| Team | 25 |
With the defaults (CSV, all tables), the output is:
# Table 0: Downloads
file,size
app-linux.tar.gz,12 MB
app-macos.zip,14 MB
# Table 1: Plans
plan,seats
Free,3
Team,25
Switch Which tables to 1 and Output format to json and you get just
that table's rows:
[
{
"plan": "Free",
"seats": "3"
},
{
"plan": "Team",
"seats": "25"
}
]
Options
- Output format —
csv(default),json,jsonl, orlist.csvwrites one delimited block per table, separated by a blank line.jsongives an array of row objects for a single table, or an array of{index, heading, line, columns, rows}envelopes when several are exported.jsonlgives one JSON value per data row; with several tables each line is wrapped as{"table": n, "row": …}.listgives an inventory only — index, nearest preceding heading, source line, column names, column alignments and row count — with no cell data.
- Which tables —
all(default), a single 0-based index like2, or a comma-separated list/range like0,2-3. Runlistfirst if you're not sure. - First row is a header — on by default. CSV keeps it as the first line and JSON/JSON Lines key each row object by it. Off, the header row is dropped and JSON emits plain arrays of values.
- CSV delimiter / quoting / line endings — a single character or
comma/tab/semicolon/pipe/space; quote only when needed (default) or every field; LF (default) or CRLF for Windows and Excel. - Trim padding inside cells — on by default, so the spaces authors use to align columns don't end up in your data.
- Render cell Markdown as plain text — off by default (cells come out
exactly as written). On,
**bold**becomesbold,`code`becomescode,[text](url)becomestext,<br>becomes a space, and\|-style escapes are resolved. - JSON indent —
0–8spaces;0minifies to a single line. - Label each CSV block — on by default; adds a
# Table n: headingcomment above each block when more than one table is exported.
Limits and edge cases
- Documents up to 1,000,000 bytes (about 1 MB of Markdown).
- Rows with fewer cells than the header are padded with empty values; cells past the header count are dropped — exactly what a Markdown renderer shows.
- A table with a header and separator row but no data rows still exports (you get the header line, or an empty array).
\|inside a cell is kept as a literal pipe, and\\as a literal backslash.- A pipe line with no separator row under it is prose, not a table, and is ignored. If nothing in the document qualifies you get an explicit error saying what a table needs, rather than empty output.
FAQ
My document has several tables — how do I get just one?
Set Which tables to its 0-based index: 0 is the first table in the
document, 1 the second, and so on. You can also pass a list or range such as
0,2-3. Pick list as the output format first to see every table's index,
heading and column names, so you know which number you want. An index past the
last table returns an error naming the valid range instead of silently giving
you the wrong data.
Why is the table in my code block being ignored?
That's deliberate. Documentation is full of fenced snippets that contain pipes —
shell pipelines, SQL, ASCII art — and treating them as data produces garbage.
Anything between ``` or ~~~ fences is skipped. If you actually want that
table extracted, remove the fence around it, or paste just the table on its own.
What happens to bold text, code spans and links inside cells?
By default nothing — cells come out byte-for-byte as written, so
[Guide](https://example.com) stays as that whole string. That's the lossless
choice. Turn on Render cell Markdown as plain text to get reading text
instead: bold and italic markers are removed, code spans are unwrapped, a link
becomes just its label, <br> becomes a space, and runs of whitespace collapse.
Underscores inside words (snake_case) are left alone.
What's the difference between JSON and JSON Lines here?
JSON gives you one document: an array of row objects when you export a single
table, or an array of table envelopes (index, heading, source line, columns,
rows) when you export several. JSON Lines gives you one compact JSON value per
line, which is what log pipelines and tools like jq -c prefer; when more than
one table is exported each line is wrapped as {"table": n, "row": …} so you
can tell the rows apart.
Will the CSV open cleanly in Excel or Google Sheets?
Yes. Fields are quoted and escaped per RFC 4180, so commas, quotes and newlines
inside a cell survive. For Excel on Windows, switch Line endings to CRLF.
For a TSV, set the delimiter to tab. If you're exporting several tables at
once, either turn off the # Table n label lines or split the blocks apart
first — a spreadsheet import expects one table per file.
Developer & Automation Access
Run it from the terminal
Same engine as this page, headless — via the gizza CLI:
gizza tool markdown-table-extractor "## Downloads
| file | size |
| --- | ---: |
| app-linux.tar.gz | 12 MB |
| app-macos.zip | 14 MB |"New to the CLI? Get gizza →
Open it by URL
Pre-fill and auto-run this tool with query parameters — the names match the API/CLI:
https://gizza.ai/tools/markdown-table-extractor/?markdown=%23%23%20Downloads%0A%0A%7C%20file%20%7C%20size%20%7C%0A%7C%20---%20%7C%20---%3A%20%7C%0A%7C%20app-linux.tar.gz%20%7C%2012%20MB%20%7C%0A%7C%20app-macos.zip%20%7C%2014%20MB%20%7C&format=csv&table=all%2C%20or%202%2C%20or%200%2C2-3&header=true&delimiter=%2C%20or%20tab%20%2F%20semicolon%20%2F%20pipe%20%2F%20space"e=minimal&newline=lf&trim=true&strip_formatting=true&json_indent=2&labels=trueMachine-readable descriptor: tool.json — title + parameters JSON Schema for agents.