Markdown Table to JSON Converter

How to use

1. 1Paste any Markdown that contains one or more pipe tables into the input on the left. Surrounding prose, headings, and code blocks are ignored.
2. 2Pick the output shape: array of objects (keys come from the header row), array of arrays, NDJSON, or a columns object.
3. 3Choose how to handle types. Leave Numbers, true/false, and Empty/null on for typed JSON, or turn them off to keep every cell as a string.
4. 4Pick the header key style (verbatim or snake_case) and the indent (minified, 2 spaces, or 4 spaces) to match your downstream tool.
5. 5Copy the JSON or click Download to save a .json or .ndjson file. Your Markdown is never uploaded.

About this tool

Markdown Table to JSON Converter parses any GitHub-Flavored Markdown (GFM) pipe table you paste into it and emits clean JSON, ready to drop into a test fixture, a mock API response, a config file, a CMS import job, or any program that speaks JSON. The parser accepts the messy real-world tables that turn up in README files and design docs: a leading or trailing pipe is optional, cells may contain backslash-escaped pipes (\|) which decode to a literal pipe, inline HTML break tags (<br>, <br/>, <br />) decode to a newline inside the cell, the five XML entities (&lt;, &gt;, &amp;, &quot;, &apos;) and numeric forms (&#NN; and &#xHH;) decode to their source characters, backtick-wrapped inline code stays a single cell even if it contains pipes, and the alignment row of dashes and colons (the :--- / :---: / ---: separator) is recognized and consumed. The detector finds every Markdown table in the input, skips fenced code blocks so a table-shaped line inside ``` ... ``` is not picked up by mistake, and pads uneven rows so partial data still becomes complete records. Pick the JSON shape that fits your consumer: an array of objects keyed by the header cells (the canonical records shape), an array of arrays for tools that take parallel data, NDJSON for streaming pipelines and one-record-per-line ingestion (BigQuery, Postgres COPY, log shippers, jq), or a columns object whose keys are the column names and values are parallel arrays. Optional type coercion turns numeric cells into real JSON numbers (with underscore separators stripped so 1_000 becomes 1000), turns true and false into JSON booleans, and turns empty cells, null, N/A, and - into JSON null, all per-cell so a single column with mixed types still round-trips correctly. Header keys can be used verbatim (preserving spaces and casing for human-friendly output) or slugified to snake_case for keys that survive every consumer. The whole conversion runs entirely in your browser, so the Markdown you paste (drafts, internal docs, README files, proprietary tables, blog posts) is never uploaded. Useful for developers turning a README table into a test fixture, technical writers exporting documentation tables to JSON for an API, data teams pulling tabular data out of issue templates and design docs, content editors importing a Markdown-authored spec into a JSON-based CMS, and anyone who needs JSON from a table they already wrote in Markdown without copying cell-by-cell.

Free to use. Works in your browser. No signup, no login.