Most developers spend hours tuning their AI coding setup. A CLAUDE.md that explains the codebase. A few custom commands in .claude/commands. Cursor rules that stop the model from rewriting your tests. Then they get a new laptop, or open the project on a second machine, and none of it is there.
Why these files fall through the cracks
Code lives in git, so it is safe. AI config often does not:
- Personal files are gitignored on purpose.
CLAUDE.local.md, personal Cursor rules and your own commands are not meant for the whole team, so they never get committed. - Global files live in your home folder.
~/.claude/CLAUDE.mdand your user-level skills apply to every project, and they sit outside every repo. - Teams do not want your experiments. Many people keep their rules out of the shared repo until they are sure they work.
So the files that hold the most of your own thinking are the ones with no backup.
Three ways to keep them safe
1. A dotfiles repo. Copy the files into a private repo and symlink them back. It works, but you have to remember to commit, and symlinks inside other repos get messy fast.
2. A synced folder. Keep the files in Dropbox or iCloud and link to them. Easy to start, but sync tools do not know which files matter, and conflicts are silent.
3. A tool built for it. Filcaas finds the AI config in each of your projects (CLAUDE.md, .claude/, AGENTS.md, .cursor/rules and more) and syncs every save in the background. Only your AI config syncs, never your code. When two machines edit the same file, the newest edit wins and the older copy is kept.
A quick check you can do today
Run this in a project and see what is not tracked:
git status --ignored --short | grep -Ei "claude|cursor|agents"Anything that shows up is a file you would lose with your laptop.