loadout
Menu
Get loadout

Every machine,
the same skills + MCP

Sync Claude Code skills and MCP servers across every machine you work from, from a reviewed shelf. Skills work with any Agent Skills client — Codex, Cursor, Gemini CLI, Copilot; MCP is projected onto the harnesses loadout detects. Pick one, and it lands on every machine you have paired — no clone, no copy, no per-machine ritual.

npm i -g loadout-agent
1 listing1 authorany Agent Skills client100% reviewed before listing
1 approved listing
publish-to-loadoutskill

Use when publishing a skill to the loadout marketplace, filling in the publish form, choosing a listing's git URL or path inside the repository, or when a published listing installs nothing on other people's machines.

@slinmorisongithub.com/sslinNn/publish-to-loadout
claude_code

1user

Install

One install, every machine, always in sync

loadout runs a small background agent on each machine you pair. The dashboard talks to those agents over a live channel, so an install, a toggle or a removal happens on the machine itself — no git pull, no setup script, no walking over to the laptop.

Pair onceInstall with npm i -g loadout-agent, run loadout pair, enter the code at /pair, and the machine joins the fleet with a live status line.
Install onceInstall from the marketplace or an inventory screen; the broadcast reaches a named target machine and is executed and read back.
Stay in syncEvery later change reconciles the fleet — one source of truth instead of N machines drifting apart.

Read before you install

Still unclear? The sync walkthrough covers install, pair, and marketplace install; how it works covers inventory, scope, and offline catch-up.

How do I sync skills and MCP configs across multiple machines?
Pair each machine once with a one-time code, then keep loadout-agent running as a background service. From then on a private realtime channel carries installs, toggles, and restores — including marketplace skills and MCP servers — to the target machine in the same round-trip. There is no repository to pull and no setup script to re-run. Full walkthrough: how to sync skills and MCP across machines.
Where can I find MCP servers?
In the public catalog under MCP servers. Browse without an account; installing onto a paired machine needs a signed-in loadout account and a running agent on that machine. The agent projects each server onto the harnesses it detects on that machine.
Where is the marketplace for skills?
Under Skills — reviewed listings you can install onto every machine you have paired. Verified-with hubs remain at for Claude Code or for Codex if you want one client at a time.
Is loadout free?
Yes. loadout is free for now — there are no paid plans. Browsing the marketplace needs no account; pairing machines and installing from the catalog needs a free signed-in account.
Which clients does loadout work with?
Skills land in .agents/skills so any Agent Skills client can pick them up. MCP is projected onto harnesses loadout detects on the machine — Claude Code, Codex, Cursor, Gemini CLI, and GitHub Copilot. Listings still show which clients they were verified with, and you can browse those hubs at for Claude Code or for Codex.
Do I need to commit my MCP server config to git?
No. loadout does not use a repository as the transport. Your fleet's state lives in your account and reaches each machine over its own channel, so a laptop that has never cloned your dotfiles still ends up with the same loadout.
What does the loadout agent actually do on my machine?
loadout-agent installs from npm and subscribes to your private channel. It applies four kinds of command — install, toggle, remove, restore — and reports the result back over the same channel. Every command is validated against a shared schema before the agent acts on it, so nothing installs itself behind your back.
Can I install a skill on a machine that is currently offline?
Yes, with loadout-agent 0.2.0 or later. The command waits in that machine's queue and runs when its daemon reconnects — after you approve it there — and the dashboard shows where it is: queued, waiting for approval, done, or denied. Older agents only hear live commands, so for them the dashboard disables install, toggle and remove until the machine is back online. Either way the switch only moves once the machine confirms it did the work.
Is my inventory visible to anyone else?
No. Machines and installed items are scoped to your account at the database level — another account's queries never return them. The only public part is a marketplace listing you deliberately publish, and that goes through review before it joins the catalog.
How is this different from dotfiles or an Ansible playbook?
A dotfiles repo or a playbook is a push you run by hand: you pull on each machine, you re-run the script, and the fleet drifts in between. loadout keeps the fleet's state in one place and pushes each change as you make it — the machine does not have to be told to go look.