I saw some discussion about this on other social media apps and wanted to know if it has been confirmed. When is it expected to happen? Are there any updates or new information?
Does anyone know what this purple dot is?
If you use Claude Code on more than one machine (laptop, desktop, work box, a server), you've probably hit this: your CLAUDE.md rules and memory live on one machine only. Every other machine starts dumb — you re-explain your preferences and re-teach the same context again and again.
claude-autosync fixes that. It symlinks your global CLAUDE.md and per-project memory into a git repo and auto-syncs it: pull on session start, commit + push on session end. One brain, everywhere.
Privacy: the tool itself stores ZERO data. Your rules and memory live in YOUR OWN private repo — not mine, not a shared one. Machine-specific stuff (paths, hosts, secrets) goes in a local.md that is gitignored and never synced.
Security: it's safe to use even if you also work on public/open-source repos.
Per-project memory defaults to a "central" mode that never writes into the project, and the opt-in "in-project" mode is automatically refused on public repos (verified via gh) so personal notes can't leak. Sync also aborts on merge conflicts instead of committing broken files.
Agent self-install: you don't really set it up by hand. Point Claude Code at the README — there's a "For AI agents" section — and it creates your private repo, detects your OS, runs the installer, and verifies everything. No manual steps.
Worth knowing: use your OWN private repo (never public) and don't put raw secrets in synced files. The richer per-project tooling is bash-first (Windows: use WSL, symlinks need Developer Mode). It's early — skim the scripts before trusting it.
MIT licensed. Feedback welcome.
