Migrate

/migrate — seamless Claude across Anton's machines

🧒 When reporting to Anton end with a child-simple "In plain words" recap. (memory eli5-always)

This is the single command over the migration project that several sessions already built. It does NOT redesign anything — it reads MIGRATION-PLAN.md + the migration-prep\* scripts and runs the right one. Full state + decisions = memory [[machine-migration]] (READ it first each run — the project is live and moves fast).

The shape (decided, do not re-litigate): always-on Windows DESKTOP = HUB (vault + automation + RAG); this laptop + Mac Studio + MacBook = satellites. Transport: small CONFIG → git/private GitHub; huge live VAULT → Syncthing star (Introducer OFF, .git hub-only). "Seamless" = same vault + same ~/.claude config + same memory on every machine, NOT teleporting the live chat.

Working dir: $USERPROFILE/!CLAUDE-HP17 May26\ · scripts under migration-prep\ · config repo $USERPROFILE/.claude/_config-backup/.

Modes

• check (default, read-only) — status + drift. Report: is _config-backup fresh (working tree clean vs HEAD; run config_sync.ps1 if skills/memory changed since)? how many Windows scheduled tasks now vs the 16 captured in migration-prep\tasks\*.xml (drift = re-export before a move)? is the F:\ chemodan stale? is gh logged in (offsite optional)? Surface what's BLOCKED.
• pack — refresh the carry-kit right before an actual move (snapshots go stale): powershell migration-prep\config_sync.ps1 (config+memory → repo) then powershell migration-prep\build_chemodan.ps1 (→ <MIGRATION_DRIVE>\, robocopy COPY-only). Re-export tasks if drifted. ⚠️ confirm with Anton before writing F:\ (it's the seed drive).
• bootstrap — on a NEW machine: read & execute _config-backup\BOOTSTRAP-NEW-MACHINE.md (ordered A→G) or migration-prep\NEW-SESSION-RUNME.md. New machine user = Anton (%WORKDIR%) → every %USERPROFILE%\ path needs _→Anton; memory encoded-path + task XML absolute paths need the rename too. **MANDATORY config-safety step:** run powershell "$env:USERPROFILE\.claude\scripts\claude_git_backup_setup.ps1" (idempotent — git-inits the ~/.claude config repos + sets a global git identity) THEN ensure the 15-min "Claude Config Git Snapshot" task exists. Without the init, the snapshot task commits NOTHING (the 2026-06-24 hub near-loss). Canon: [[claude-skills-git-backup]] + [[config-safety-backup-and-migration-check]].
• connectors — reconnect external MCPs after a move via migration-prep\CONNECTORS-INVENTORY.md. The 3 blind spots the vault+config backup MISSES: <TELEGRAM_MCP_DIR>\ (.env + patched source), $USERPROFILE/.claude/secrets/ (n8n.env + SSH key), $USERPROFILE/.local/share/whatsapp-mcp/. Enumerate from .claude.json mcpServers, not the backup. After reconnecting, verify each server is actually alive & authed with the **/mcp** skill (its in-session health check — and it carries the Telegram AUTH_KEY_DUPLICATED landmine that bites right after a move: re-auth per machine, never copy .claude.json).
• sync — Syncthing setup/repair via migration-prep\SETUP-SYNC.md. First-sync order is CRITICAL: laptop is most-current → laptop = Send&Receive, desktop = Receive-Only on first pass to ABSORB, THEN flip. Laptop Device ID is in [[machine-migration]]. Same Syncthing pipe also carries the cross-machine RELAY — Claude-to-Claude messages between machines (folder $OBSIDIAN_VAULT/_machine-bus/); that is the /inbox skill's job, NOT /migrate (rule lives in [[reglament-multi-machine-claude-i-peredacha-mezhdu-mashinami]]).

Hard "never" (safety invariants)

• NEVER copy $USERPROFILE/.claude.json to another machine (Telegram MCP session) → AUTH_KEY_DUPLICATED logs the account out. Re-auth per machine.
• NEVER start a 2nd MCP client on the same StringSession (see [[mcp-health-check]]).
• _originals (3.75 GB) is SACRED — send-only from hub, never delete.
• All migration .ps1 are ASCII-only (PowerShell 5.1 mangles Cyrillic — [[deterministic-script-gotchas]]).
• Writes to F:\ / pushing the repo / registering tasks on a new box = pause + confirm (Tier-2, [[operating-agreement]]).

Output

One-liner status + what's fresh/stale/blocked + the exact next command. Then 🧒 recap. Don't duplicate logic that lives in the scripts — point at them.

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything is open source and free, so take what you need.

Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), telegram-mcp-kit (your own Telegram over MCP in about 15 minutes).

Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

Engineers: want to test-drive this setup? Message me. I hand out free starter seeds to engineers who test and report back, and custom skill requests are welcome.