How to convert JSON to a Markdown table

1. Drop a .json, .jsonl, .ndjson or .ldjson file on the page, or click Paste data, paste JSON or JSONL and click Open.
2. Hide the columns you don't want in the Columns panel, and narrow the rows with the filter box.
3. Click Export and choose Markdown.
4. Paste the table into your README, pull request or docs.

DuckDB-WASM reads the file inside your browser, so nothing is uploaded unless you click Share. The grid has no row limit and loads rows in pages of 100, while the Markdown export writes the first 1,000 rows of your current view.

A .json file over 1.5 GB is refused with "Too big for this browser". A larger JSONL file loads partially, with a note saying how many rows fit.

rows.page flattens nested objects into dot-path columns such as user.address.city, up to 8 levels deep and 200 child fields per object. A column holding JSON text is expanded too when 90% or more of its values in the first 1,000 rows are JSON objects.

If the file is one top-level wrapper such as {"data": [...]}, rows.page unwraps it to its biggest list, up to 4 levels deep, and shows the note: Showing the N items in "data"; the other top-level fields were skipped. The export then keeps the pipe table valid:

For the input [{"id": 1, "user": {"name": "Ada"}, "tags": ["a", "b"], "note": null}], the export looks like this:

| id | user.name | tags | note |
| --- | --- | --- | --- |
| 1 | Ada | ["a","b"] |  |

Line breaks inside a value become spaces, because a newline would end the row. The delimiter row is always plain ---, so add colons yourself if you want columns aligned.

Export writes your current view: the filter, sort order and visible columns. If the JSON has dozens of nested keys that would make a README table too wide, hide the extra columns first. To cut rows without SQL, type in the filter box or click histogram bars and top values in the Columns panel:

• country:NL matches a value.
• price:>20 compares numbers.
• name:* keeps rows where the field has a value.
• -name:* keeps rows where the field is empty.

To give readers every row, click Share to upload the file and get a https://rows.page/<id> link they can filter and sort next to the 1,000-row table in your pull request.

To keep a few fields or rename dot-path headers, type SQL into the filter box. A statement that starts with SELECT, WITH or FROM runs against the table data, and column names with dots need double quotes:

SELECT id, "user.name" AS name FROM data

rows.page uses DuckDB SQL, so window functions, QUALIFY, unnest(), PIVOT inside a subquery and FROM (SUMMARIZE data) work. A hidden __rid column holds the row number: SELECT DISTINCT * EXCLUDE (__rid) FROM data removes duplicates, and ORDER BY __rid keeps the file order.

SQL results show up to 1,000,000 rows, and CSV, JSON, JSONL or Parquet exports of a result stop there. A Markdown export of a result still stops at 1,000 rows.

In a script or notebook, flatten the records with pd.json_normalize() and call .to_markdown():

import json
import pandas as pd

with open("data.json") as f:
    records = json.load(f)

print(pd.json_normalize(records).to_markdown(index=False))

.to_markdown() needs the tabulate package and raises an import error without it. pandas has no 1,000-row cap, but lists print as Python lists such as ['a', 'b'], a | inside a value is not escaped, and you write the code to filter rows or unwrap a top-level object like {"data": [...]}.

jq has no Markdown formatter, so you list the fields and join them with pipes yourself:

jq -r '
  ["id", "user.name"],
  ["---", "---"],
  (.[] | [.id, .user.name])
  | "| " + join(" | ") + " |"
' data.json

This suits small, predictable files in a pipeline. join turns nulls into empty cells but fails on arrays and objects, so wrap those fields in tojson, and it does not escape a | inside a value.