I tried using tmuxp last week and there was a very bizarre lack that completely killed its use for me. Specifically, while you can set a start directory for the session and for the window, nowhere in the docs could I find a way to set the start directory for a pane. And that's a problem because I really don't want all of my panes to be set up with a visible `cd path/to/folder` at the top.
So I ended up just writing a quick shell script that set up my environment using regular tmux commands.
This sounds so useful in theory (I tried it) but I think it just doesn't work with my flow of operations. I'm mostly anti-mouse and pro-keyboard but somehow this combination of xmonad window layouting and using terminator (where I exclusively rearrange my split shells with the mouse) has been useful for years and while I'm a thankful and frequent user of tmux on servers I never could be productive on my desktop... :(
And it's been doing the trick for years. Prior to building the tmux workspace, it even can call a bootstrap script (https://github.com/tony/tmuxp/blob/master/bootstrap_env.py) to check for and install dependencies.
When a tmuxp session file (https://github.com/tony/tmuxp/blob/master/.tmuxp.yaml) is loaded, it automatically can source the user into virtualenvs, launch file watchers, and so on.
Saves so much time.
Since most software projects involve the same flow (installing packages locally for the project, launching a test runner.)
Other things that help keeping it cross platform are using Makefile and commands conforming to POSIX (https://github.com/tony/tmuxp/blob/master/Makefile). So when bringing up a list of files for entr (a file watcher http://entrproject.org/) to run commands when files are saved, use posix find arguments only. Use $(MAKE) for recursive calls (https://www.gnu.org/software/make/manual/html_node/Recursion...).