MWoffliner is a tool for creating a local offline HTML snapshot of any online MediaWiki instance. It scrapes all pages (or a selection if specified) and creates the corresponding ZIM file. While primarily targeted for Wikimedia projects like Wikipedia and Wiktionary, MWoffliner also supports any recent MediaWiki instance (version 1.27+), though instances with custom skins or highly unusual configurations may have limitations.
Read CONTRIBUTING.md to learn more about MWoffliner development.
User help is available in the FAQ.
- Scrape with or without image thumbnails
- Scrape with or without audio/video multimedia content
- S3 cache (optional)
- Image size optimization and WebP conversion
- Scrape all pages in namespaces or title list based
- Specify additional/non-main namespaces to scrape
Run mwoffliner --help to see all available options.
- Docker (or Docker-based engine)
- amd64 or arm64 architecture
The recommended way to install and run mwoffliner is using the pre-built Docker container:
docker pull ghcr.io/openzim/mwofflinerRun software locally / Build from source
-
*NIX Operating System (GNU/Linux, macOS, etc.)
-
Redis — in-memory data store
-
Node.js version 24 (we support only one single Node.js version; other versions might work or might not)
-
Libzim — C++ library for creating ZIM files (automatically downloaded on GNU/Linux & macOS)
-
Various build tools which are probably already installed on your machine:
libjpeg-dev— JPEG image processinglibglu1— OpenGL utility libraryautoconf— automatic configuration systemautomake— Makefile generatorgcc— C compiler
(These packages are for Debian/Ubuntu systems)
An online MediaWiki instance with its API available.
-
Clone the repository locally:
git clone https://github.com/openzim/mwoffliner.git && cd mwoffliner
-
Build the image:
docker build . -f docker/Dockerfile -t ghcr.io/openzim/mwoffliner
[!WARNING] Local installation requires several system dependencies (see above). Using the Docker image is strongly recommended to avoid setup issues.
Setting up MWoffliner locally for development can be tricky due to several dependencies and version requirements. Follow these steps carefully to avoid common errors.
MWoffliner requires Node.js 24 (other versions may fail).
Compatible Node 24 ranges: >=24 <24.6 or >=24.7 <25.
Check your version:
node -vIf your version does not match, use nvm to install the correct Node.js version.
MWoffliner depends on @openzim/libzim, which requires the C++ libzim library.
- On Linux/macOS, MWoffliner can download libzim automatically.
- On Windows, you must install libzim manually because there are no prebuilt binaries. See the libzim installation guide for details.
Node 24 on Windows officially supports Visual Studio 2019 (v16) or Visual Studio 2022 (v17).
Ensure C++ build tools are installed and environment variables are set correctly. See Windows Setup for node-gyp for detailed instructions.
MWoffliner uses node-gyp, which enforces strict checks for Node and compiler versions. Make sure you have:
- Proper Visual Studio version (Windows) — see Visual Studio versions
- Required C++ headers, e.g.,
zim/archive.h— see libzim documentation - Python 3.10+ (required by node-gyp; a recent version is preferred for compatibility)
-
Clear npm cache — a corrupted cache can cause cryptic install failures:
npm cache clean --force
-
Delete node_modules and reinstall — stale or partially installed dependencies are a common source of errors:
rm -rf node_modules package-lock.json npm install
-
Check that all environment variables are set — especially on Windows,
PATH,INCLUDE, andLIBmust point to the correct Visual Studio and libzim directories. Reopen your terminal after installing new tools. -
Verify Redis is running before starting MWoffliner — MWoffliner will fail immediately if it cannot connect to Redis:
redis-cli ping # expected output: PONG -
Run npm install with verbose logging to see exactly where it fails:
npm install --verbose
| Error | Cause | Solution |
|---|---|---|
| Node.js version error | Node.js version incompatible | Install Node 24 with nvm |
| Cannot find module @openzim/libzim | libzim not installed | Follow libzim installation guide; Windows users must install manually |
| node-gyp rebuild failed | Wrong Node or compiler version | Check Node.js version, Visual Studio version, Python 3.x |
| zim/archive.h not found | C++ headers missing | Install libzim system-wide, verify include paths |
[!NOTE] Even with these steps, other setup errors may occur. Using Docker is strongly recommended for a smoother experience.
npm i -g mwoffliner[!WARNING] You might need to run this command with the
sudocommand, depending on how yournpm/ OS is configured.npmpermission checking can be a bit annoying for newcomers. Please read the npm script documentation if you encounter issues.
# Get help
docker run -v $(pwd)/out:/out -ti ghcr.io/openzim/mwoffliner mwoffliner --help# Create a ZIM for https://bm.wikipedia.org
docker run -v $(pwd)/out:/out -ti ghcr.io/openzim/mwoffliner \
mwoffliner --mwUrl=https://bm.wikipedia.org --adminEmail=foo@bar.netUsing NPM / Local Install
# Get help
mwoffliner --help# Create a ZIM for https://bm.wikipedia.org
mwoffliner --mwUrl=https://bm.wikipedia.org --adminEmail=foo@bar.netTo use MWoffliner with an S3 cache, provide an S3 URL:
--optimisationCacheUrl="https://wasabisys.com/?bucketName=my-bucket&keyId=my-key-id&secretAccessKey=my-sac"If you've retrieved the MWoffliner source code (e.g., via a git clone), you can install and run it locally with your modifications:
npm i
npm run mwoffliner -- --helpDetailed contribution documentation and guidelines are available.
MWoffliner provides an API and can be used as a Node.js library. Here's a stub example for your index.mjs file:
import * as mwoffliner from 'mwoffliner'
const parameters = {
mwUrl: 'https://es.wikipedia.org',
adminEmail: 'foo@bar.net',
verbose: true,
format: 'nopic',
pageList: './pageList',
}
mwoffliner.execute(parameters) // returns a PromiseWarning
MathJax support is UNSTABLE: the --mathJax* CLI parameters described below may change, even in a minor release. Setting them up also requires wiki-specific technical preparation, so this is aimed at developers rather than end users.
MWoffliner can bundle MathJax into the ZIM so that math formulas keep rendering offline, on wikis that rely on it (e.g. via the SimpleMathJax extension).
MathJax 2, 3 and 4 are very different beasts: each ships its own set of files and, more importantly, each requires its own incompatible configuration format (which defines things like math delimiters and custom macros). Because of this, MWoffliner cannot auto-detect and configure MathJax for you — you have to supply a matching MathJax build and configuration yourself, via three CLI parameters:
--mathJaxSource: local path or HTTP(S) URL to a ZIP archive of a compiled MathJax distribution. Its content is extracted and pushed to the ZIM (under an internal_mathjax_/namespace).--mathJaxConfig: local path or HTTP(S) URL to an HTML file containing a single<script>tag with the MathJax configuration (this must be copied from the wiki, see below). Its content is injected inline, before the MathJax library, on every page that needs it. If the configuration needs to reference a path inside the MathJax archive (e.g.MathJax.Ajax.config.path[...]), it cannot use an absolute path since ZIMs have no fixed root URL. Write__MATHJAX_ROOT__instead of the leading slash and mwoffliner will replace it, on every page, with the correct relative path to the root of the MathJax archive you provided. For instance useMathJax.Ajax.config.path["Contrib"] = "__MATHJAX_ROOT__/MathJaxExtensions/legacy";if the resources are located in aMathJaxExtensions/legacyfolder in the archive (inside the root folder of the archive if there is a single folder as usual).--mathJaxEntryPoint: path, relative to the root of the extracted archive, to the MathJax script to load (e.g.es5/tex-chtml.jsfor MathJax 3). Defaults toMathJax.js(MathJax 2 entry point).--mathJaxAllPages: inject the config/entry-point<script>tags on every page instead of only on pages detected to need MathJax (see below). Some wikis don't list MathJax in their page JS modules, which defeats the automatic detection; this flag works around that at the cost of adding the scripts to every page.
--mathJaxConfig, --mathJaxEntryPoint and --mathJaxAllPages all require --mathJaxSource to also be set. A page is considered to "need" MathJax, and only then gets the config/entry-point <script> tags injected, when one of the JS modules MediaWiki reports for that page matches mathjax (case-insensitive), unless --mathJaxAllPages is set, in which case every page gets them. The extracted library files themselves are always pushed to the ZIM as soon as --mathJaxSource is set, regardless of which pages use them.
- Find the live MathJax version and configuration. Open a wiki page that renders math formulas, open your browser's developer console and type
MathJax.versionto get the exact version. Then find the configuration, typically a<script>block settingwindow.MathJax = {...}(MathJax 3/4) orMathJax.Hub.Config({...})(MathJax 2) in the page source — copy it as-is into a local file, e.g.mathjax-config.html. - Find the exact entry point. In your browser's network tab, find the request loading the MathJax library itself (typically named
MathJax.jsfor MathJax 2, or atex-chtml.js/tex-svg.js/... for MathJax 3/4) and note its full path, including any query string. - Build a matching MathJax ZIP, the exact steps depend on the major version in use (see below).
- Run the scraper with the three parameters, e.g.:
mwoffliner --mwUrl=https://your.wiki --adminEmail=foo@bar.net \ --mathJaxSource=./mathjax-source.zip \ --mathJaxConfig=./mathjax-config.html \ --mathJaxEntryPoint=es5/tex-chtml.js
MathJax 2 is not published to npm as an installable package; it is only distributed as source on GitHub. Download a release archive directly from the MathJax releases page (e.g. 2.7.9) and use it as-is as --mathJaxSource, no re-zipping needed:
curl -Lo mathjax-source.zip https://github.com/mathjax/MathJax/archive/refs/tags/2.7.9.zipThe archive has a single top-level folder (e.g. MathJax-2.7.9/); mwoffliner strips it automatically when extracting.
Unlike MathJax 3/4, MathJax 2 loads its extensions/output-processor via a config= query parameter on the MathJax.js request itself (e.g. MathJax.js?config=TeX-MML-AM_CHTML) rather than solely through the injected configuration script — this is the URL you captured in step 2 above. Set --mathJaxEntryPoint to that same path and query string, e.g.:
--mathJaxEntryPoint="MathJax.js?config=TeX-MML-AM_CHTML"The referenced combined-configuration file (here config/TeX-MML-AM_CHTML.js) is part of the standard MathJax 2 distribution, so it is already included in the ZIP from the release archive.
MathJax 3 is published to npm as mathjax-full, which bundles both the compiled runtime (under es5/) and the TypeScript sources (which mwoffliner automatically ignores when an es5/ folder is present):
npm install mathjax-full@3.2.2
cd node_modules/mathjax-full && zip -r ../../mathjax-source.zip . && cd ../..Entry point example: --mathJaxEntryPoint=es5/tex-chtml.js.
MathJax 4 moved to scoped npm packages. Use the plain mathjax package (deployment-ready bundle), not @mathjax/src (the TypeScript source package meant for building MathJax itself, which requires separately installing font packages):
npm install mathjax@4
cd node_modules/mathjax && zip -r ../../mathjax-source.zip . && cd ../..Unlike MathJax 3, entry-point files live at the root of the package (no es5/ folder), e.g. --mathJaxEntryPoint=tex-chtml.js.
Complementary information about MWoffliner:
- MediaWiki software is used by thousands of wikis, the most famous ones being the Wikimedia ones, including Wikipedia.
- MediaWiki is a PHP wiki runtime engine.
- Wikitext is the markup language that MediaWiki uses.
- MediaWiki parser converts Wikitext to HTML, which displays in your browser.
- Read the scraper functional architecture for more details.
GPLv3 or later, see LICENSE for more details.
This project received funding through NGI Zero Core, a fund established by NLnet with financial support from the European Commission's Next Generation Internet program. Learn more at the NLnet project page.


