How to sync Claude Code skills and MCP configs across machines

Install the loadout agent, pair each machine, then install Claude Code skills and MCP servers from the marketplace so they land on every paired machine over a live channel.

1. npm i -g loadout-agent

   Install the loadout agent. On each machine you want in the fleet. Needs Node 20 or newer — see Install for platforms and the background service. The package installs loadout and loadout-agent — both point at the same binary.

2. loadout pair

   Pair the machine. Prints a 12-character code and waits. Enter it at /pair (or open the URL the command printed). Then run loadout service install so the agent keeps running after you log out — full detail on Pair a machine. Repeat on every laptop or desktop you want in sync.

3. Install from the marketplace. Browse skills or MCP servers, then install a listing onto a paired machine. The dashboard addresses that machine’s agent over a private realtime channel — the install lands on disk there in the same round-trip. No git pull and no setup script to re-run.

4. Stay in sync across the fleet. Later toggles, removals, and restores follow the same path: one change in the dashboard, applied by each target machine’s agent. Offline machines catch up when their daemon reconnects — see How it works. Inventory for a machine reflects what is actually on disk there.

Confirm it worked

Each paired machine turns online in Machines once its agent is running and has sent a heartbeat. After an install, that machine’s row in Inventory lists the skill or MCP server that landed on disk — not a plan for what should be there.

What syncs (and what does not)

loadout syncs the skills and MCP servers you install, toggle, remove, or restore through the dashboard — tagged for claude_code or codex, scoped global or project. It is not a general dotfiles sync and does not invent shared team or org fleets: each account’s machines and inventory stay scoped to that account.