Pandoc turns 20 in 2026 — what every Markdown writer should know

On August 3, 2026, Linuxiac reported that Pandoc — the open-source tool used everywhere to convert one markup format into another — was celebrating its twentieth birthday. The initial release landed on August 3, 2006, under the GPL license, and the project has been quietly powering publishing pipelines ever since.

If you write in Markdown, Pandoc is probably already part of your life even if you have never run it. Its HTML output shapes how thousands of static sites and blogs are built. Its DOCX and EPUB writers turn plain-text notes into finished documents that open anywhere. And its new ability to run in a web browser points directly at the same privacy-first, no-upload workflow that md2rich is built around.

This guide breaks down what Pandoc is, the milestone the 2026 anniversary marks, the brand-new in-browser capability that changes how you can use it, and — most practically — when you should reach for Pandoc versus a lightweight client-side converter.

A quick history: from a Haskell learning project to a publishing staple

Pandoc began, the way many beloved tools do, as a personal project. Its creator John MacFarlane, a philosophy professor and mathematician, built it in Haskell partly to learn the language. The very first version, Pandoc 0.1, weighed in at about 3,000 lines of code and could convert documents between Markdown, reStructuredText, HTML, and LaTeX, while also generating RTF and S5 slide-show output.

That modest start solved a problem most writers still have today: you almost never want just one output. You draft in a readable syntax, but your editor wants a Word file, your publisher wants LaTeX or EPUB, your website wants HTML, and your slides want a presentation format. Pandoc's answer was to build a single engine with pluggable readers and writers rather than a pile of bespoke converters.

Because a reader converts any input format into a common abstract syntax tree, and a writer renders that tree into any output format, Pandoc does not need a separate tool for each possible format pair. Add one reader and one writer and a format can take part in many combinations at once. That architectural bet — now twenty years old — is the reason the project has stayed so flexible without collapsing under its own weight.

The 2026 retrospective put real numbers on that design: more than 200 versions released across two decades, four main packages containing over 85,000 lines of Haskell (not counting tests), more than 600 contributors, and 7,346 GitHub issues closed. The latest release at the time of writing is 3.10.1.

What Pandoc actually is

People often call Pandoc a document converter, but it is more precisely a programmable document-processing system. Instead of keeping a separate converter for every pair of formats, Pandoc uses two kinds of components: readers, which parse an input format into a common abstract syntax tree, and writers, which render that internal representation into the desired output. Because every reader and writer plugs into the same tree, any format that has a reader can be combined with any format that has a writer.

That architecture is why the format list keeps growing without the project ballooning in complexity. According to the 2026 retrospective, Pandoc now supports a large set of markup, word-processing, academic, presentation, and publishing formats — including Markdown, HTML, LaTeX, DOCX, ODT, EPUB, AsciiDoc, reStructuredText, Org Mode, Typst, Jupyter Notebook, PowerPoint, DocBook, and JATS, along with a number of wiki and bibliography formats.

Pandoc has also moved well beyond plain text conversion. It handles citations and bibliographies, mathematical expressions, templates and reference documents, presentations, and electronic books. In other words, it is a foundation for automated publishing workflows — exactly the kind of pipeline we covered in our round-up of the 2026 Markdown tool landscape.

Two more features make Pandoc more than a batch file converter. First, filters: you can alter the structure of the intermediate document with custom filters that run against the abstract syntax tree, transforming headings, images, or metadata however you like. Second, Lua scripting is built in, so you can write small scripts that reshape documents without recompiling Pandoc. For developers this turns Pandoc into a genuine programmable publishing layer, not merely a converter.

Twenty years in numbers

The 20th-anniversary retrospective by Pandoc's creator, John MacFarlane, lays out the scale of the project:

What started as a personal project to learn Haskell became one of that language's best-known applications. That trajectory mirrors the wider story of Markdown tooling: small, local, plain-text tools that outgrow their humble beginnings because they solve a real problem predictably.

The big 2026 change: Pandoc in the browser

The most notable new capability in recent Pandoc releases is the ability to compile the application to WebAssembly. Pandoc 3.9 introduced this as a major feature, and it changes what "universal document converter" can mean.

With the WebAssembly build, a largely fully featured version of Pandoc runs directly in a web browser — without sending your documents to a remote server. The conversion happens on your device. That is the same privacy-first promise that makes client-side conversion attractive for anyone who handles sensitive or unpublished writing.

MacFarlane's own words on the AI question are worth reading. He acknowledges that AI systems can already translate between some document formats, but he maintains that Pandoc keeps several advantages: much lower energy requirements, predictable output, and more reliable conversions. When you need a deterministic file to hand to a publisher or an academic journal, you do not want a model improvising the layout.

A quick Pandoc workflow with real examples

Pandoc is both a command-line program and a Haskell library. The most common way to use it is the pandoc command. Here is how a typical Markdown-to-Word conversion looks:

# Convert Markdown to a Word document
pandoc post.md -o post.docx

# Convert Markdown to a self-contained HTML file
pandoc post.md -s -o post.html

# Convert Markdown to EPUB (an e-book)
pandoc post.md -o post.epub

# Render a bibliography from a Markdown + BibTeX source
pandoc paper.md --citeproc --bibliography=refs.bib -o paper.pdf

Because Pandoc reads and writes from your existing files, it fits beautifully into scripts and automated pipelines. You can bundle every article in a folder, convert the whole set to DOCX for an editor, to EPUB for an e-reader, and to PDF for archival — all from one source of truth.

Here is a concrete Markdown fragment and what it becomes under Pandoc's standard writer. Your Markdown source:

# Launch announcement

**Date:** 2026-08-06

Today we published the **20th-anniversary edition** of the 
guide. Highlights:

- Local, offline conversion
- No upload, no tracking
- Predictable output

> Convert once, publish everywhere.

Feed that to pandoc launch.md -o launch.html and Pandoc writes the equivalent structured HTML: an <h1> for the title, a paragraph with the date, a <ul> for the list, and a <blockquote> for the quoted line. The same source can ship as a polished .docx for a stakeholder or an .epub for readers.

Pandoc lets you specify input and output formats explicitly with the -f (from) and -t (to) flags, though it usually infers them from file extensions. Those flags are especially useful when you want a format that is not obvious from a filename:

# Convert an Org Mode file to a Markdown file explicitly
pandoc notes.org -f org -t markdown -o notes.md

# Convert Markdown to a PDF via LaTeX (needs a LaTeX engine)
pandoc report.md -o report.pdf --pdf-engine=xelatex

# Convert a folder of Markdown files to a combined EPUB
pandoc chapter-*.md -o book.epub --toc

Getting started: how to install Pandoc in 2026

Installing Pandoc depends on your operating system, and the options are broad enough that practically everyone can run it:

• Windows — a self-contained installer is available, and Pandoc also ships in Windows Package Manager (winget install JohnMacFarlane.Pandoc) and Chocolatey.
• macOS — a .pkg installer, plus Homebrew (brew install pandoc).
• Linux — packages in most distributions (apt install pandoc on Debian/Ubuntu, dnf install pandoc on Fedora), or a portable binary.
• In the browser — the WebAssembly build added in Pandoc 3.9 means you can run a largely fully featured Pandoc without installing anything, on any device with a modern browser.

Once installed, pandoc --version prints the version you are running, and pandoc --list-input-formats and pandoc --list-output-formats show every format your build can read and write. The WebAssembly edition brings that same capability to people who would rather not touch a terminal.

A privacy-minded take on the browser build

The WebAssembly milestone is significant beyond convenience. When a converter compiles to WebAssembly and runs in a browser, the document no longer has to cross the network to be transformed. For writers handling drafts, contracts, unpublished research, or any material that should not sit on a third-party server, that is a meaningful privacy guarantee — the same zero-upload principle behind client-side Markdown converters.

MacFarlane frames the durability of Pandoc partly in these terms. In discussing AI-based format translation, he notes that while AI systems can already translate between certain document formats, Pandoc retains clear advantages: much lower energy requirements, predictable output, and more reliable conversions. When you are shipping a manuscript or a paper, "predictable" is worth a great deal — you want the same input to produce the same output every single time, not a plausible-sounding variation.

The energy point matters too. A locally compiled converter that turns a few files into different formats costs almost nothing to run. The same task on a remote AI endpoint involves inference, network transfer, and electricity on someone else's servers — and it sends your private document along with it.

When to reach for Pandoc — and when to reach for md2rich

Pandoc is extraordinarily capable, but it is also a heavyweight generalist. There are two very different jobs hiding inside "document conversion." It is worth choosing the right tool for each.

Pandoc is the right choice when you are building automated pipelines, working with academic or technical formats (LaTeX, JATS, bibliography tools), generating e-books, or converting between many formats at once. It rewards people who script their publishing. It is free, open source, offline, and deterministic.

md2rich is the right choice when the task is narrower and more common: you write Markdown and want to paste it as formatted rich text into LinkedIn, X Articles, Medium, or Notion. You do not want to install anything, learn a command line, or manage a LaTeX toolchain. A single-page app that converts in your browser, with nothing uploaded and no account, is the fastest path from draft to published post.

You can see this distinction in practice across our Markdown to X Articles guide and the Markdown editor round-up. Different jobs call for different tools, and the best workflow often uses both.

The larger lesson: local, predictable, yours

Pandoc's twentieth birthday is a useful reminder of what plain-text publishing tools get right. Twenty years in, the tool still works offline, produces predictable output, and leaves you in control of your files. Those are the same values that drive the privacy-first publishing movement we have covered in articles like the State of Markdown Publishing round-up and the privacy-first Fediverse workflow.

Whether you use a powerful converter for your archive and a lightweight one for your daily posting, the principle stays the same: keep the canonical source on your device, convert client-side, and never hand your unpublished writing to a server you do not control. On its twentieth anniversary, that is the strongest argument for Pandoc — and for every tool that follows in the same spirit.

FAQ