How to connect Wolfram Engine to Jupyter on Ubuntu?

It has been reported that the official Jupyter client from Wolfram works fine on Windows and Mac at the moment. However, some have gotten it to work under Linux by supplying the right path argument to the kernel as in this GitHub ticket. See cc-wr's comment below.

An alternative iPython/Jupyter notebook "kernel" called JWLS should work out of the box

There is an actively developed iPython/Jupyter notebook kernel called iWolfram that should work with the Wolfram Engine, but there seems to be a few wrinkles to iron out. This GitHub ticket tracks progress on iWolfram support for Wolfram Engine.

In the mean time, you can still call Mathematica from Python code as you always have been able to do: How to use Mathematica functions in Python programs?.

This answer is a community wiki which should be updated by anyone as more details are available.

I followed the instructions at https://github.com/WolframResearch/WolframLanguageForJupyter/blob/master/README.md

wolframscript configure-jupyter.wls add

but I get

configure-jupyter.wls: Provided Wolfram Engine binary path does not exist.

This issue should be fixed with the most recent commit.

so I tried

wolframscript configure-jupyter.wls add ["/usr/local/Wolfram/WolframEngine/12.0/SystemFiles/Kernel/Binaries/Linux-x86-64/WolframKernel"]

where the path shown is the result of starting wolframscript and doing

First[$ComamndLine]

and get the same result.

You do not need the square brackets.

Also, you may run into an issue with your use of the ".../SystemFiles/Kernel/Binaries/..." WolframKernel path: it may not be meant to be called directly by the user.

Instead, you should use the "canonical" WolframKernel path in Executables. An example is below:

/usr/local/Wolfram/WolframEngine/12.0/Executables/WolframKernel

If you are looking for something that just works give a try to https://github.com/Ludwiggle/JWLS If your wolframscript works, JWLS works too, cause JWLS is a bash kernel that reads and writes to fifos; as such, it hardly fails.