Sync ~/.claude/CLAUDE.md user memory across user's machines (multi-device single user)

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

When using Claude Code on multiple machines as a single user (desktop + laptops + servers), each machine has its own ~/.claude/CLAUDE.md user memory. The user's identity/preferences/context don't follow them across machines.

Result: a new session on machine B starts blank about user habits already encoded on machine A. Cousin Claude doesn't know who the user is, what fleet conventions apply, or what the user has been working on elsewhere.

Concrete example from my own setup (5-machine fleet, single user sky1241):

On sky-master (primary machine), Claude has full context loaded from ~/.claude/CLAUDE.md:

• "Sky lives in Switzerland (Europe/Zurich)"
• "Sky's REAL portfolio: MUNINN, yggdrasil, bible-ux..."
• "Don't push toward all-nighters / exhausting sessions..."
• (etc., dozens of personal preferences and project context)

Same user on pc3 (a fleet laptop), Claude session:

User: "bonjours bonsoir sinon tu sais qui je suis ?" Claude: "Non, j'ai rien sur toi en mémoire — première session. Je sais juste : ton email c'est ludovik1241@gmail.com, et tu bosses sur bible-ux. Dis-moi qui t'es et ce qu'on attaque, je note."

Same user, same Anthropic account, same gh auth login sky1241 → completely blank Claude. The 5+ minutes user-context onboarding has to repeat each time the user moves machines.

In another session on pc3, Claude failed to retrieve a file via SSH. It found wg0 10.13.13.4/24 (an active VPN interface), but stopped at "~/.ssh/config is empty → no access". It never tested the VPN connection because it had no memory that the user runs a WireGuard fleet. The Claude on sky-master would have known immediately.

Proposed Solution

Any one of these solves it (pick whichever fits the stack):

• A. Cloud sync of ~/.claude/CLAUDE.md tied to the Anthropic account, like VS Code Settings Sync
• B. CLI commands claude memory export <file> / claude memory import <file> with simple JSON
• C. "Load memory from URL" config option pointing to a user-controlled gist/repo (no Anthropic infrastructure needed)

Alternative Solutions

Manual git-tracked dotfiles repo with ~/.claude/CLAUDE.md symlinked. Works but requires Linux dotfiles culture, not discoverable.

Third-party tools (e.g. my own MUNINN sub-token compression) can fill the gap but add a layer that wouldn't be needed if Claude Code shipped with simple sync.

Priority

Low — nice to have. Affects only multi-device single users.

Feature Category

Configuration and settings

Use Case Example

Single user, 5 Linux machines (Ryzen desktop + 2 laptops + Pi server + mini-PC), all logged in as the same Anthropic account via gh auth login. Fleet shares: WireGuard VPN, mining cluster, Telegram bot, AdGuard DNS, backup chain. User context is identical on all 5 machines, but Claude memory is siloed per-host.

Each new fleet machine integration = manual rsync ~/.claude/CLAUDE.md between hosts, or symlinks to a git-tracked source. Adds friction. Cloud sync (Option A) would remove it.