markitdown

markitdown

A pure-Go library and CLI that converts documents to Markdown. Go port of the Python markitdown library.

Features

• Pure Go, no CGO, no external runtime dependencies
• 12 format converters: PDF, DOCX, PPTX, XLSX, XLS, HTML, RSS/Atom, CSV, EPUB, Jupyter, plain text, ZIP
• Deterministic output with golden test suite
• PDF extraction via PDFium (WebAssembly, no CGO) with heading/bold/italic detection

Supported formats

Install

go get github.com/conductor-oss/markitdown

Library quick start

package main

import (
	"fmt"
	"log"

	markitdown "github.com/conductor-oss/markitdown"
)

func main() {
	m := markitdown.New()

	// Convert a local file
	result, err := m.ConvertFile("report.pdf")
	if err != nil {
		log.Fatal(err)
	}
	fmt.Println(result.Markdown)
}

More examples

// Convert a URL
result, err := m.ConvertURL("https://example.com/page.html")

// Convert with auto-detection (file path or URL)
result, err := m.Convert("report.pdf")
result, err := m.Convert("https://example.com/page.html")

// Convert from a reader with metadata hints
f, _ := os.Open("data.csv")
result, err := m.ConvertReader(f, markitdown.StreamInfo{
	Extension: ".csv",
	MIMEType:  "text/csv",
	Charset:   "shift_jis",
})

// Options
m := markitdown.New(
	markitdown.WithKeepDataURIs(true), // preserve base64 data URIs in output
)

CLI quick start

Build:

go build -o markitdown ./cmd/markitdown

Convert a file to stdout:

./markitdown report.pdf

Convert and write to a file:

./markitdown -o output.md report.docx

Convert from stdin with format hint:

cat data.csv | ./markitdown -x csv

Convert a URL:

./markitdown https://example.com/page.html

CLI flags

Usage: markitdown [flags] [source]

Arguments:
  source    File path or URL to convert (reads stdin if omitted)

Flags:
  -o, --output string       Output file (default: stdout)
  -x, --extension string    File extension hint for stdin input (e.g. "pdf", ".csv")
  -m, --mime-type string    MIME type hint
  -c, --charset string      Charset hint (e.g. "shift_jis", "utf-8")
  -v, --version             Show version
      --keep-data-uris      Keep full base64-encoded data URIs in output

Notes

• PDF extraction is text-based; image-only PDFs produce no output without OCR.
• DOCX math equations (OMML) are converted to LaTeX notation.
• CJK charset detection works without hints but is most reliable when Charset is provided in StreamInfo.

Acknowledgements

This project is a Go port of Microsoft's markitdown Python library. The original project provides the reference implementation, test fixtures, and design that this port is based on.