Convert HTML to PDF in C# — without a headless browser. Pure managed code: no native binaries, no browser, no LibreOffice, no Office interop. Runs on Linux, macOS, Windows and arm64, and makes no network calls at runtime.
| From ↓ / To → | CSV | DOC | DOCX | HTML | Markdown | PPTX | XLSX | |
|---|---|---|---|---|---|---|---|---|
| CSV | — | · | · | · | · | · | · | · |
| DOC | · | — | ✅ | · | · | · | · | · |
| DOCX | · | · | — | ✅ | ✅ | ✅ | · | · |
| HTML | · | · | ✅ | — | · | ✅ | · | · |
| Markdown | · | · | ✅ | · | — | ✅ | · | · |
| · | · | · | · | · | — | · | · | |
| PPTX | · | · | · | · | · | ✅ | — | · |
| XLSX | ✅ | · | · | ✅ | · | ✅ | · | — |
A ✅ is a converter that ships; · is a pair with no converter, not a promise about one. Read a row as "from this format, into these".
Generated from the shipped API, not written by hand — full grid, plus every edit/read operation. Also: open, edit and password-protect DOCX, XLSX and PPTX, and read PDF text.
dotnet add package Ank.DocToolkit// HTML in, a Word document and a PDF out. No browser, no LibreOffice, nothing to install.
byte[] docx = await HtmlToDocxConverter.ConvertAsync("<h1>Invoice 2026-114</h1>");
byte[] pdf = await HtmlToPdfConverter.ConvertAsync("<h1>Invoice 2026-114</h1>");
// Read a document back, edit one you were given, and lock it when you are done.
string text = DocxEditor.ExtractText(docx);
byte[] locked = PdfEditor.Protect(pdf, new PdfProtection { UserPassword = "s3cret" });Targets net8.0 and net10.0. MIT licensed.
📖 Guides · 🔎 API reference · 🗺️ Roadmap · 📦 package README · 📝 CHANGELOG
Contributing? Branch from
mainand open a pull request back into it —mainitself cannot be pushed directly. See CONTRIBUTING.md for the build, the commit-message rules, and the four constraints that will get a pull request rejected.
Most .NET document stacks fail one of three things: a licence with a revenue threshold, a native binary that breaks on Linux, or a claim nobody re-checked. This package is built to fail none of them, and the numbers below are how that gets verified rather than asserted.
xychart-beta
title "runtimes/ folder in a plain build (MB)"
x-axis ["Ank.DocToolkit", "EPPlus", "QuestPDF", "NPOI", "ShapeCrawler"]
y-axis "MB in build output" 0 --> 700
bar [0, 1, 83, 416, 664]
Those bars are the runtimes/ column below, which is not the same measurement as the native-file
count beside it: EPPlus's 1 MB contains zero .so/.dylib files. Both are in the table.
| package | native .so/.dylib in build output |
runtimes/ |
licence in NuGet metadata |
|---|---|---|---|
| Ank.DocToolkit | 0 | 0 MB | MIT, as an SPDX expression |
| EPPlus | 0 | 1 MB | ships license.md — read it |
| QuestPDF | 10 | 83 MB | ships LICENSE.md — read it |
| NPOI | 12 | 416 MB | ships OSMFEULA.txt — read it |
| ShapeCrawler | 19 | 664 MB | none declared |
Every number was measured on 2026-08-09 by adding the package to an empty net8.0 console app and
building it. Reproduce any row in under a minute — that is the point of publishing the method
rather than a conclusion:
dotnet new console && dotnet add package <name> && dotnet build -c Release
find bin -path '*runtimes*' \( -name '*.so' -o -name '*.dylib' \) | wc -l
du -sm bin/Release/net8.0/runtimesTwo things this table deliberately does not do. It does not tell you what those licences say — they are linked from each package's own page and are the authors' to describe, not ours. And it does not claim native payload is everyone's problem: EPPlus carries essentially none, so if that is your only concern it is not a reason to switch. Where the payload does appear it comes from SkiaSharp and Magick.NET, pulled in transitively for image rendering.
What the table is for is the case where both columns matter at once — a Linux container that has to stay small, with a licence you can clear without asking anybody. That combination is the whole reason this package exists, and all four constraints below are re-checked by CI on every push.
| Convert | HTML and Markdown to DOCX and PDF; DOCX to HTML, Markdown and PDF; XLSX to CSV, HTML and PDF; PPTX to PDF; legacy Word 97-2003 .doc to DOCX; legacy PowerPoint 97-2003 .ppt to PDF (at a measured rate) |
| Edit | create and edit DOCX, XLSX and PPTX; fill templates, including one row per record; insert images; read text back out |
| page count, merge, split, extract, rotate, reorder, insert; read text; read and stamp document information | |
| Protect | password-protect and unprotect PDF, DOCX, XLSX and PPTX |
The full grid is generated from the shipped API rather than written by hand: what it can convert.
Task-shaped walkthroughs, one per format, start at Getting started for the install-to-first-conversion path, or jump straight to the page for the format you need.
Most .NET document stacks fail at least one of these. This one satisfies all four:
| Constraint | How |
|---|---|
| Free for commercial use | 35 dependencies: 33 MIT, 2 Apache-2.0. No revenue thresholds, no per-seat fees. |
| NuGet only | No Chromium download, no LibreOffice install, no native binaries. |
Targets net8.0 and net10.0 |
Two LTS targets, one public API surface. net9.0 is deliberately absent — a net9.0 app already consumes the net8.0 build, so it would add no reach. netstandard2.0 is deliberately absent too: every dependency supports it, but the bounded-fetch guarantee on remote images cannot be expressed there, and DateOnly/TimeOnly would make the API differ per target. |
| Runs everywhere .NET does | The full suite runs in CI on Linux, Windows, macOS and arm64 (ubuntu-24.04 x64, windows-latest, macos-latest Apple Silicon, ubuntu-24.04-arm). Not inferred from "pure managed" - measured on each. |
| Works offline | No runtime network I/O. Proven by 40 air-gap tests. |
All four are properties of the resolved dependency graph, so a single upstream bump can break them silently — which is why CI re-checks every one on every push.
That has happened once already. PowerPoint support originally used ShapeCrawler, whose API
was checked and whose dependencies were not. It pulls in SkiaSharp and Magick.NET, which put
38 native .so/.dylib files and 664 MB of runtimes/ into build output — breaking two of
the four constraints at once, on a package that restored and built perfectly well. It was replaced
with raw DocumentFormat.OpenXml, and the check that would have caught it now runs on every push.
The comparison table above is the same measurement, applied
to the alternatives.
| Trimmable / native AOT | Both packages marked, and both run-tested in CI — not just published warning-free. One caveat: ClosedXML emits trim warning IL2090. → Trimming and native AOT |
| Provenance attestation | Every .nupkg and SBOM is signed, naming the workflow, commit and runner. → Verifying a release |
| SBOM | CycloneDX per package, attached to every GitHub Release |
| Offline, enforced not intended | 40 guard tests assert zero socket connections against 16 ways markup can name a host; mutation-tested weekly. → Nothing reaches the network unless you ask |
| Fidelity, on real files | 88–98% success across 500+ real .gov documents, monthly. → Measured on documents nobody here wrote |
| What this won't do | CSS layout, mixed page setups, multi-line headers — the bounded list. → Known limitations |
Offline is the default everywhere; remote images are the one opt-in, and it stays bounded rather than wide open:
byte[] docx = await HtmlToDocxConverter.ConvertAsync(html, allowRemoteImageDownload: true);
// Bounded instead of wide open: timeout, byte cap, host allow-list and a block on
// loopback/private/link-local addresses, all on by default. Not a complete SSRF defence — see
// the package README.
byte[] bounded = await HtmlToDocxConverter.ConvertAsync(html, new RemoteImageOptions());The package README is the reference: every capability, every overload, and the caveats that matter. It is also what nuget.org renders, so it is what you get on the package page.
| Guides | Task-shaped walkthroughs - getting started, HTML, Markdown, Word, spreadsheets, DI, production |
| API reference | Generated from the source |
| Samples | One runnable project per capability, each answering one question, built against the published packages |
| Package README | Full usage, page setup, passwords, telemetry, known limitations, and the migration notes per release |
| Extensions README | services.AddDocToolkit() and the injectable interfaces |
| CHANGELOG | What changed, per release |
| PUBLIC.md | Why the design docs, ADRs and backlog are not in your clone, and what would have to change for them to be |
| CONTRIBUTING | Build, test, commit rules, repository layout, and the four constraints a pull request must not break |
| SECURITY | Reporting a vulnerability, the scope, and the documented limits that are not findings |
MIT — see LICENSE.
If this replaced a headless browser in your build, a star helps others find it — GitHub's repository search ranks partly on stars, and this one has three.