The hub for AI Skills

Publish, discover, and use AI agent skills across all platforms.
Connect the MCP server once and every skill on the registry becomes instantly available — no local installation needed.

💡 Why iKhono?

AI coding agents are powerful but generic. Skills give them specialized expertise — security review, test writing, API design — crafted by the community.

• 🔌 Use a skill: Connect the MCP server and search for what you need
• ✏️ Create a skill: Write a SKILL.md, publish with one command
• 🌍 Share a skill: Your skills work on Claude, Cursor, Windsurf, Copilot, and Codex

🚀 Quick Start

1. Add the MCP Server

Recommended (stdio — full featured, auto-reads auth token):

npm install -g @ikhono/cli
ikh setup --platform claude

Or manually: claude mcp add ikhono -- npx -y @ikhono/mcp

Zero-install (Streamable HTTP — great for trying it out):

claude mcp add ikhono https://ikhono.io/mcp -t http

Search and load skills work without auth. For pins and ratings, pass -H "Authorization: Bearer YOUR_TOKEN".

Cursor / Windsurf / Copilot: See platform setup guides.

2. Use Skills

Once connected, your AI agent can search and load skills automatically. Just ask:

"Search iKhono for a code review skill"

Or the agent will suggest skills when it detects a relevant task.

3. Explore

Action	Command	What Happens
Search	/skill search	Find skills by query, category, or author
Load	/skill load	Load full instructions into your AI agent
Pin	/skill pin	Save favorites for quick access
Unpin	/skill unpin	Remove a skill from favorites
List Pinned	/skill list-pinned	Show your pinned skills
Rate	/skill rate	Help the community find the best skills

🔍 How It Works

You  -▶  AI Agent  -▶  iKhono MCP Server  -▶  iKhono Registry
                              ⬇
                            search
                              ⬇
                    Skills loaded on-demand
                      into your AI agent

The MCP server is a thin proxy — it connects your AI agent to the iKhono registry. When you (or your agent) search for a skill, the instructions are fetched and injected into the conversation. No files downloaded, no local setup.

🖥 Supported Platforms

Platform	Setup Guide	Status
Claude Code	.claude/settings.json	Supported
Cursor	.cursor/mcp.json	Supported
Windsurf	.windsurf/mcp.json	Supported
GitHub Copilot	.vscode/mcp.json	Supported
OpenAI Codex	Environment variables	Supported
Claude Desktop	claude_desktop_config.json	Supported
Gemini CLI	~/.gemini/settings.json	Supported

🛠 For Skill Creators

Have expertise to share? Publish a skill in minutes:

# Install the CLI
npm install -g @ikhono/cli

# Login
ikh login              # opens browser for GitHub SSO

# Scaffold a new skill
ikh skill init my-skill

# Edit skill.yaml and SKILL.md, then publish
ikh skill publish --changelog "Initial release"

Your skill is now available to everyone on the registry as @your-username/my-skill.

Resources

• Creating Skills Guide — Full walkthrough
• Skill Format Spec — skill.yaml and SKILL.md reference
• CLI Reference — All commands and options
• Example Skills — Skills you can reference and fork

📚 Documentation

Doc	Description
Getting Started	5-minute setup and first skill
How It Works	Architecture and concepts
Creating Skills	Authoring and publishing guide
Skill Spec	Format reference
CLI Reference	Command-line tool docs
Platform Guides	Per-platform setup instructions

🤝 Contributing

See CONTRIBUTING.md for how to contribute skills and help test on different platforms.

❓ FAQ

• What does iKhono mean?
  iKhono means "skill" in Zulu.

• Do I need to install each skill separately?
  No. The MCP server gives you access to the entire registry.
  Skills are loaded on-demand when you or your AI agent requests them.

• Is it free?
  Yes. Publishing and using skills is free.

• Can I use private skills?
  Not yet. All published skills are public. Private/team skills are on the roadmap.

• What platforms are supported?
  Any AI tool that supports the Model Context Protocol (MCP).
  We have setup guides for Claude Code, Cursor, Windsurf, GitHub Copilot, OpenAI Codex, Claude Desktop, and Gemini CLI.

📄 License

Documentation and examples in this repository are licensed under MIT.