mm-wiki: frozen by its author, and four ports for one service
MM-Wiki is a lightweight enterprise knowledge-sharing and team collaboration tool for quickly building a corporate wiki and team knowledge-sharing platform. Easy to deploy and use, it helps teams build a collaborative environment for information sharing and document management.
At a glance
- What is it?
- MM-Wiki is a Chinese-language enterprise wiki written in Go and shipped as a per-platform binary. Its readme opens with an unusually candid note from the author saying he stopped maintaining it because both the interface and the technology stack needed a rewrite, and it is that same readme that names four different ports for the same service across its install, container and proxy instructions.
- Who is it for?
- MM-Wiki fits a team that wants a self-hosted wiki in one binary, with spaces, roles, external login and search already arranged, and that does not need a modern interface. Two things to weigh before you deploy it.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Activity is slowing. The repository last received commits 7 months ago.
- What is it written in?
- Mainly Go, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- phachon/mm-wiki3,781 ★ · +5
| Repository | 2026-09-24 | 2026-11-15 |
|---|---|---|
| phachon/mm-wiki | 3.8k | 3.8k |
5 Stars in the last 52 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
The author froze it, in writing, and said why
The first thing in this readme is not a feature list. It is the maintainer apologising.
He thanks the people using it, says that seeing their messages makes him feel ashamed, and that he has not had much energy for maintenance in recent years. Then the reason: when he was iterating on new features he found that both the interface and the technology stack would need a thorough rewrite. So he stopped and left the current version as it is.
He adds that he has decided to start planning an entirely new version, and invites people to join in. And he says why he built it: five years ago he wanted it to be the best free open-source wiki available domestically, something as great as Confluence.
That is a rare thing to find at the top of a project page. Most abandoned projects simply stop being updated. This one says which of the two reasons applies, which is that the code is not the problem, and it names the two things that are.
It also sets expectations honestly. Anyone evaluating this for a new deployment should read it as a frozen release that is documented rather than a project with a roadmap.
Four ports, one service, and they do not agree
The install section is the most useful part of the readme and also the most confusing, for one reason: the port changes four times.
The bundled installer listens on 8090 by default, and the readme tells you to override it with a flag if you want a different one:
$ ./installThen the container section, where a comment says the service starts on port 8080, and the command beside it maps host 8080 to container 8081. Those two statements are already inconsistent with each other.
Then the optional reverse proxy example, whose upstream block points at the service on 127.0.0.1:8088, with a comment noting that is where the application listens.
So the same service is documented as listening on 8090, on 8080, on 8081 and on 8088 within about forty lines of instructions.
This is not a criticism so much as a description of what happens when a document is maintained across years without anyone reconfiguring an example. But it is the first thing to check when you install, because the failure mode is a connection refused on the port you expected.
The container instructions are a patch behind the newest release
The Docker section offers two routes, and the numbers in them are old.
The first route is to pull a published image, and the version named is 0.1.7, which the readme marks as the recommended option. The second is to build the image from the current source locally and run that.
The newest release in the repository's tag list is 0.2.1. So the container path everyone is told to take installs a version two patch releases behind what the project has released, and the alternative is to build from source yourself, which undoes the convenience of a container.
The setup is otherwise sensible: mount a configuration directory and a data directory, and point the service at its configuration file.
The database step is the one that contradicts the rest of the product's pitch. Elsewhere the readme advertises an installer with a browser-based setup and no manual steps. The container instructions require you to import two SQL files from the documentation directory, and one of those files needs a statement uncommented by hand before the administrator account exists.
So the container route is the one place in this project where you touch a SQL file, which is worth knowing before you pick it.
Three Go versions in one repository
The development section says one thing about the environment, the module file says another, and the container says a third.
The readme states the requirement as Go 1.8, with a clone, a directory change and a build as the whole sequence.
The module file declares a language version of 1.12, and the module dependencies include a web framework from that era.
The container file builds with a 1.14.1 toolchain image.
Three versions, six minor releases apart at most, in three places a contributor will look in order. The container is the one that works today; the readme is the one a new contributor reads first.
The build itself is minimal in the readme and more elaborate in the container, where the source is cloned from the forge at image build time rather than copied from a local directory, the timezone is set explicitly, and there are commented-out lines showing how to point the module fetcher at a regional mirror if the default one is slow.
Those commented lines are a small time capsule: they exist because building this image from outside the author's region was a real problem, and the fix was documented rather than assumed.
The plugin list dates the interface
There is a section thanking the plugins the interface is built from, and it is the most revealing page in the readme.
It lists fourteen front-end libraries: a CSS framework and its satellite plugins for checkboxes, an icon picker, select boxes, switches and tag inputs; a browser markdown editor; a layout helper; a layer system; a collapsible menu; a charting library; a popover; a scroll helper; and a tree view.
Every one of those is a jQuery plugin from the period when jQuery plugins were distributed this way. Which is the evidence behind the maintainer's own statement that the interface needs a thorough rewrite: he was not being pessimistic about a detail, he was describing a front end assembled from libraries that have not been the default choice for years.
It also predicts what the planned rewrite means. Not a reskin. If the interface is rebuilt on a current component approach, the whole of that list is replaced, and the features built on top of it, spaces and permissions and search, are what carry over.
So for anyone deciding whether to wait for the new version, that section is the best available evidence of how much work is planned.
The dependency list shows a framework of that era, and a few leftovers
The module file is short and it reads like a snapshot of a particular year.
The web framework is Beego, at a version from that period. There is an LDAP client, which is how the readme's external login feature is implemented. There is a MySQL driver, an object-relational mapper, a validation library, a mail library for the notification feature, and a Markdown renderer for the document format.
There is also a process-metrics library, and a terminal colour stack: a colour package, a colourable wrapper, a Windows-specific variant, and an isatty check. Those are for the installer's console output rather than for the web interface.
Two entries are worth naming because nothing in the readme suggests them. A Windows Management Instrumentation package, marked as indirect, which is the kind of thing a process library pulls in for one platform. And an application-engine package from a different era entirely, also indirect.
Then three test libraries, all marked indirect, which is unusual enough to suggest the test suite either lives elsewhere or does not run in the module's own build.
Taken together, this is what a long-lived Go web application looks like: the framework and helpers that were current when it was written, plus the sediment of things that arrived through dependencies and never left.
A public demo with its credentials in the readme
There is an online demo section, and it contains a working address, a username and a password.
For a demo instance that is a reasonable thing to publish: the point is to let someone log in and look around without installing anything, and a readme where the credentials are hidden helps nobody.
It is still worth naming as a pattern, because the demo and a real deployment are the same binary with the same configuration format. There is nothing in the readme telling you to change anything after the first login.
The rest of the readme closes the way a Chinese open-source project usually does: an invitation to buy the author a tea, an official group chat number, and a request for stars, issues, pull requests, forks and email.
There is also a contributors section listing five people, three of whom are credited with code and two with documentation, one of them credited with both. That table is generated from a configuration file at the repository root, which means adding a contributor is a pull request like any other.
Sources at the root, views beside them, and logs committed
The layout is worth one paragraph because it tells you what kind of application this is.
Two Go files sit at the repository root rather than in a command directory, alongside the module files and a router file. Then the application directory, the view templates, the static assets, the configuration directory, an install directory, a logs directory, and the documentation directory with the database scripts the container instructions tell you to import.
Two shell scripts for building and packaging sit at the root as well.
Two more root entries are unusual. There is an English version of the readme next to the Chinese one, which is rare for a project with a Chinese-first document and a small number of international contributors. And there is a site configuration file for a static site generator, in a repository that builds a single binary, which suggests the project page was generated rather than served by the application.
The logs directory is committed, and that is the one entry with a security consequence rather than a cosmetic one: a directory that exists to receive runtime output, tracked in version control, is a directory that will eventually receive something sensitive.
Editorial conclusion
MM-Wiki fits a team that wants a self-hosted wiki in one binary, with spaces, roles, external login and search already arranged, and that does not need a modern interface. Two things to weigh before you deploy it. The project is frozen by its author's own statement, so what you get is the last version he was willing to ship rather than a maintained line, and the newest release tag is six years old. And the documentation has drifted: the container instructions are a patch behind the newest release, three different Go versions appear across the readme, the module file and the container, and the port you are told to use changes four times between the installer, the container and the proxy example. None of that makes it unusable. It does mean you should read the whole install section before you trust any single line of it.
Frequently asked questions
What is MM-Wiki?
A lightweight enterprise wiki and team collaboration tool written in Go and distributed as one binary per platform. It provides spaces as permission-scoped collections of documents, Markdown authoring with attachments, custom roles with per-role permissions, external login such as LDAP, email notifications when a document changes, document sharing and download, and full-text search.
How do I install MM-Wiki?
Download a prebuilt archive for your platform, unpack it, and run the bundled installer, which serves a browser-based setup page on port 8090 by default. Stop the installer, then start the binary with a configuration file. Separate steps are given for Linux and Windows, followed by an optional reverse proxy example.
Is MM-Wiki still being updated?
The readme carries a note from the author saying it has not been actively maintained in recent years because both the interface and the technology stack needed a thorough rewrite, that the current version is intentionally left as it is, and that a new version is planned with community involvement. The newest release tag is from September 2020.
What database does MM-Wiki use?
MySQL, with the connection details set in the configuration file. The container instructions tell you to import two SQL files from the documentation directory, with a note that you must uncomment the first statement in one of them to create the administrator user.
What is the license for MM-Wiki?
The MIT license, stated at the end of the readme and with a license file at the repository root.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/phachon-mm-wiki)Star history chart for your README
[](https://hysenlabs.com/projects/phachon-mm-wiki)Was this useful?