sx CLI
The sx CLI and app keep a team's skills, rules and MCP in a shared vault and install them into everyone's AI clients
- 315 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- Writes skills, rules, hooks and MCP servers into the settings of several AI clients
- Installed hooks and MCP entries are run by the client later and need the sx binary
- The cloud relay exposes the vault through an MCP address for claude.ai and chatgpt.com
Install
- Works in
Manual install
brew install sleuth-io/tap/sxmacOS and Linux. Then sx init in the vault or project. The app is downloaded separately from releases.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Writes skills, rules, hooks and MCP servers into the settings of several AI clients
- Installed hooks and MCP entries are run by the client later and need the sx binary
- The cloud relay exposes the vault through an MCP address for claude.ai and chatgpt.com
This is third-party code. Review the repository files before installing.
What it does
sx is a team's AI asset manager, covering everything from skills and rules to hooks, MCP servers and Claude Code plugins. It keeps them in a shared vault with versions and installs them into the right clients in their native formats. A vault can be a folder the team already syncs (for example through Dropbox or Google Drive), a git repository or the Skills.new cloud. A folder needs no server or accounts. sx has two front doors with one shared configuration. In the desktop app for macOS, Windows and Linux a skill is created in a built-in editor or dragged in as a file. The CLI is for engineers and automation: `sx add` puts an asset into the vault and detects its type, and `sx install` installs it by scope, from the whole organization down to one repository, user or bot (for example a CI runner). `sx stats` shows usage and token spend, `sx audit` keeps a change log, and `sx vault copy` moves a vault between types. Through the Skills.new relay the vault is exposed as an MCP address for claude.ai and chatgpt.com.
Who it is for. For teams where several people use different AI clients and want to share skills and rules without copying between repositories.
Good fit when
- Skills live in one repository but are needed in others, and copies drift apart
- Non-technical colleagues should also publish and get skills without git: a shared Dropbox or Drive folder is enough
- One asset must go into several clients at once, for example Claude Code, Cursor and Codex
- Assets must reach only the right audience: one team, one repository or a CI bot
- You need usage stats and a change log: who installed what and who changed it
Not a fit when
- A single developer just wants to sync skills on one machine: a local manager such as skillshare is simpler
- You need change requests with roles and approvals: that mode (RBAC) is on the roadmap and not ready
- You want assets inside a Git repo with no separate vault: sx assumes a vault
- You need stable MCP server support: that asset type is marked experimental
Example request
Add my code-review skill to the vault, install it for the whole organization, then show usage stats for the last weekLimitations
The MCP servers asset type is marked experimental. Client support is uneven: Claude Code gets all asset types, Gemini in JetBrains gets only rules and MCP, Android Studio only remote MCP over HTTP, and KiroCrew only skills with global scope. The app is downloaded separately from releases and ships its own CLI copy that updates with the app and does not self-update. Hooks committed to a repository (Copilot, Kiro) need `sx` on PATH.
How to disable. Remove installed assets with sx for the relevant scope and remove the CLI itself with `brew uninstall sx` or by deleting the binary. The app is removed like any other program. The relay connection ends when you stop `sx cloud serve`.
FAQ
How do I install sx?
On macOS and Linux: `brew tap sleuth-io/tap` and `brew install sx`, or the script `curl -fsSL https://raw.githubusercontent.com/sleuth-io/sx/main/install.sh | bash`. Then `sx init` in the vault or project. The app is downloaded from releases (`sx-app-*` artifacts).
Do I need a server or account?
Not for a folder vault: any directory the team already syncs will do, for example `sx init --type path --path ~/Dropbox/sx-vault`. Git vaults and the Skills.new cloud are also available.
How do I limit who gets a skill?
Choose the scope on install: `--org` for everyone in the vault, `--repo github.com/acme/infra` for one repository, `--path` for a directory inside it, `--team platform`, `--user alice@acme.com` or `--bot python-backend` for a bot.
Which clients are supported?
Claude Code, Cline, Codex, Cursor, GitHub Copilot, Gemini (CLI and VS Code, JetBrains, Android Studio), Kiro, KiroCrew, Openclaw and OpenCode.
How do I see what is actually used?
`sx stats` shows an adoption dashboard, and `sx stats --since 7d --json` gives a machine-readable report for the last week. `sx audit` reads the log of team and install changes.
Discussion
No questions or reviews yet.
Related
CLI toolsA desktop app and CLI to install, sync and organize agent skills across Claude Code, Codex, Cursor and other tools
An MCP server that opens SKILL.md skills to clients without native support, such as Codex, Copilot and Cursor
A CLI and MCP server for Agent Skills: spec validation, installs from GitHub and on-demand loading of the right skill
A CLI skill manager for AI agents: install from repositories, translate between agent formats and deploy into their configs
skills (npx skills)
skills
A skill package manager: installs, finds, updates and removes SKILL.md packages from GitHub and other git hosts across many agents
A zero-dependency Python CLI: one catalog of MCP servers and skills for every agent, with only a compact index in context
README in short
The README opens with a short walkthrough for the app and for the CLI and explains when each front door fits better. Scoped distribution, audit, a client table and portable bot agents follow.