Clyx (pronounced like clicks) is a concatenative, stack-oriented, dynamically-typed, interpreted programming language designed for simplicity, minimalism, and metaprogramming capabilities.

Clyx aims for maximum portability and can be implemented in any modern runtime environment that supports data structures defined in JSON (numbers, strings, arrays and key-value dictionaries). Extra features varying in portability include filesystem support, external environment calls and TCP socket I/O.

The core spec of Clyx is mostly stable but the standard library is still a work-in-progress.

The most recent Clyx version is 0.3. The version is bumped every time the language core words (aka primitives), the standard library or the REPL get updated. The primitive set is expected to be indefinitely frozen once the version number reaches 1.0.

Language features

Single, unbounded data stack that can hold numbers, strings and lists (here called Q-forms)
Forth-like parsing, RPN (postfix notation) and word definition mechanics
Tcl-like bracket syntax and Janet-like list operations on Q-forms
Homoiconicity: Q-forms and strings can be manipulated as data and executed dynamically
Character I/O (implementation-dependent)
File I/O (may not support all features on some targets)
TCP socket I/O (may not be supported on some targets)
Host language interop (may not be supported on some targets, only to be used when nothing else helps)
Platform-agnostic standard library
Ability to include custom modules into the current runtime context (depends on file I/O capabilities)
A simple REPL (written in Clyx itself) shipped with almost every port.

Implementations

Reference implementation in Python (compatible with MicroPython and T-DeckARD on CircuitPython)
Reference implementation in Go

Installation

Depending on the implementation, there are several installation ways.

Reference Python (CPython) implementation installation

For the easiest way to install the reference Python-based Clyx implementation, pipx is recommended:

pipx install "git+https://codeberg.org/luxferre/clyx.git#subdirectory=clyx-python"

This will install the REPL as clyx into your current environment.

Then, you can just run pipx upgrade clyx to upgrade to the new release versions whenever they come out.

Reference implementation for MicroPython

Clone this repo, then run sh dist-python.sh. A dist/clyx_py/clyx directory will appear, which will contain all the files necessary to run micropython clyx.py [prog.clx] on the desktop or copy them to the target device.

Reference implementation for CircuitPython + T-DeckARD runtime

Clone this repo, then mount your T-Deck's CircuitPython volume, and then run:

sh dist-python.sh /path/to/mounted/volume/app/clyx

This will install Clyx as a T-DeckARD applet.

Reference Go implementation installation

To install the Go-based Clyx implementation and its tools (including clyxc) directly from Git, run:

go install codeberg.org/luxferre/clyx/...@latest

(Note: If you are installing immediately after a new commit or release has been pushed, Go's default module proxy might serve a cached version. You can bypass this cache by running GOPROXY=direct go install codeberg.org/luxferre/clyx/...@latest instead).

For maximum performance, you can pass optimization flags to build stripped binaries without debug symbols, bounds checks, or inlining:

go install -a -trimpath -gcflags=all="-l -B -e" -ldflags "-s -w" codeberg.org/luxferre/clyx/...@latest

This will install the executables as clyx and clyxc into your $GOPATH/bin directory. To upgrade to the latest versions at any time, simply run the same command again.

Alternatively, to compile and build it locally, navigate to the clyx-go directory and run:

cd clyx-go
make

This will build the clyx executable.

Usage

Note: this section describes how to run the interpreter itself. For the language manual, refer to Clyx Reference Manual.

All core implementations must accept one or more .clx files to run (either via command line or via library call parameter). If multiple files are passed, they are executed in sequence. If a _main word is defined in the workspace after running all source files, it is automatically executed as the entry point. The REPL is implemented in Clyx itself and resides in the clyx_repl.clx file.

Running the reference Python implementation

If Clyx is installed as a PyPi package, use clyx [prog1.clx prog2.clx ...] to run the REPL or one or more programs.

Otherwise, to run the REPL, run [micro]python clyx.py command from the directory created by ./dist-python.sh. Pass one or more .clx files additionally to run them directly.

Running as a T-DeckARD applet

To run the REPL inside T-DeckARD:

from app.clyx import clyx
clyx.repl()

To run a Clyx source file directly, issue clyx.clyx('path/to/file.clx') instead.

Caveats:

The entire operation is rather slow (which is expected for an interpreted runtime on top of CircuitPython on this class of machines) and may sometimes overflow the allocated (native) Python stack. As such, it's not recommended to src files from the REPL if you can run them directly instead.
Don't forget to mount an SD card or run unlock_rootfs() if you want to use file writing capabilities inside T-DeckARD environment.

Running the Go implementation

To run the REPL, execute the compiled clyx binary without arguments:

./clyx

To run programs directly, pass the path to the .clx files as arguments:

./clyx path/to/program1.clx path/to/program2.clx

Self-contained Packaging (Compilation)

The Go implementation supports reading an optional ZIP archive appended directly to the end of the clyx binary. This allows you to package your Clyx source code into a single executable file.

To package your program, you can use the clyxc utility.

Installing `clyxc`

Install the utility directly from Git (this also installs the main clyx binary):

go install codeberg.org/luxferre/clyx/...@latest

Or to install only the compiler utility:

go install codeberg.org/luxferre/clyx/clyxc@latest

Alternatively, if compiling locally from the repository root:

go install ./clyxc

Using `clyxc`

Run the compiler by passing the output binary path followed by the list of source files or directories:

clyxc out-bin-path src1 src2...

(Note: clyxc requires the main clyx interpreter binary to be installed in your $GOPATH/bin directory).

Manual Packaging Alternative

If you prefer not to use clyxc, you can package the application manually:

Create a ZIP archive containing all your .clx source files.

Append the ZIP archive to the compiled clyx binary:

cat clyx my_app.zip > my_app_executable
chmod +x my_app_executable

When the packed executable runs:

The interpreter automatically detects the appended ZIP archive.
It iterates through all .clx files in the archive and evaluates (src) them.
If a _main word is defined in the environment, it is executed automatically.
If _main is not present, the interpreter falls back to running the standard REPL.

FAQ

Where does the name Clyx originate from?

Similarly to LVTL and LTTB, the prototype of Clyx had been called LPLF (Luxferre's Pointer-Less Forth). However, as the finalized language began to differ from the Forth standard in too many ways, a decision was made to focus on the concatenative nature itself. Thus, the word ConCat had been chosen, which was quickly modded into ConLynx and then naturally shortened to Clyx.

Additionally, this name can be viewed as an acronym hinting at the order some binary operations pop their operands from the stack: "Compute by Loading Y, then X".

Which language(s) does Clyx look like the most?

While working in a totally different manner than either of those, Clyx was obviously heavily inspired by some design ideas of Forth and Tcl.

Is this language high-level or low-level?

More of a high-level one (because you don't have any memory control), but in fact it can be both.

Clyx combines the conceptual simplicity of Forth (and concatenative/RPN programming in general) with list definition syntax, manipulation and metaprogramming capabilities of Janet and Tcl, making it much more practical for everyday tasks. At the same time, you can build abstraction levels as high as you need to, making it a nice choice for creating custom DSLs (domain-specific languages), especially with the next word that pushes the next logical token onto the stack, allowing to change token processing order.

Why not just pick up a Forth instead?

Forth is good if you are fine with only being able to push integers onto the stack and building all your abstractions using direct memory access. Clyx, on the other hand, does not have direct memory access because it was designed for the systems that might not even give you one, but you can do so much more with the data types that it allows you to push onto the stack without involving any trickery.

While diverging from some "purity" aspects, Clyx is a decent choice for newcomers that want to try out a totally new programming paradigm without investing too much time into fighting the language's quirks.

Why are lists called Q-forms in Clyx?

Idiomatically, lists (as a data structure, as in Python lists or JSON arrays) can only hold data. Q-form is short for "quotation form", which just means that it is a quoted (block-like) sequence that can be treated both as data and as code. Physically, Q-forms are implemented as random-access lists or arrays in 100% cases, but logically, it's important to maintain this distinction that Q-forms can hold program code and lists in general do not. Besides, the term "list" is confusing in general because, apart from the random access lists used to implement Q-forms, it can also refer to linked lists or double-linked lists, which have totally different sets of features and constraints.

What do I need to do to create my own Clyx implementation?

Just port the core interpreter along with the 50 primitive words, stubbing out the words that cannot be supported by the target platform if there are any, so that if a program that e.g. uses nopen runs on a non-networked host environment, then the user would get a clear message that an unsupported core primitive word has been encountered and the program has been terminated because of that.

Apart from that, your implementation needs to be able to load lib.clx (standard library source) upon initialization, and clyx_repl.clx upon passing no parameters (if it is CLI-based).

Doesn't `hseval` primitive make Clyx programs non-portable across different implementations?

Yes, it does, because it works differently across different host environments and languages (e.g. Python expression evaluation vs. Go system shell command execution), so it must only be used as a last resort when any other attempts to interact with the outside world are unsuccessful. Additionally, using hseval introduces extra security risks. Nevertheless, this primitive had been added in order to enable some host-specific interoperability that would otherwise be impossible or very cumbersome to implement, such as browser-based deployments.

In other words, hseval is sometimes very useful in very specific cases, but please avoid using it unless you have a really good reason to use it.

Credits

Created by Luxferre in 2026, released into the public domain with no warranties.

README.md

Clyx programming language

About