I built xeus-haskell: a lightweight Haskell kernel for Jupyter (and it runs on JupyterLite!)
I’ve been playing with MicroHs, a wonderfully minimal Haskell implementation, and it inspired me to build a new Jupyter kernel: xeus-haskell.
A few fun things about it:
-
It’s built on MicroHs, so it has almost zero dependencies.
-
Because of that minimalism, it compiles cleanly to WebAssembly.
-
Which means… you can run Haskell in the browser via JupyterLite.
-
No GHC, no giant toolchains, no environment wrangling. Just a browser.
The goal is to make Haskell more accessible in scientific/technical computing. Lazy evaluation can be surprisingly powerful for graph algorithms, recursive structures, and anything where “compute only what’s needed” brings real wins. Being able to demo that interactively in a notebook feels like the right direction.
If you want to check it out:
Repo: https://github.com/tani/xeus-haskell
Demo (JupyterLite): https://tani.github.io/xeus-haskell
Feedback, suggestions, and wild experiments welcome!