home / mcp / skills mcp server
Bring Claude's Skills pattern to any MCP-compatible agent
Configuration
View docs{
"mcpServers": {
"skills-mcp-skills-mcp": {
"command": "npx",
"args": [
"-y",
"skills-mcp",
"-s",
"/absolute/path/to/skills"
]
}
}
}Transform your AI agent into a domain expert by attaching Skills through the Model Context Protocol. This MCP server exposes reusable, modular skills in Claude’s format so your agent can discover and apply them across tools and platforms with a simple, centralized configuration.
You can use the Skills MCP server with any MCP-compatible agent. Once the server is running, your agent will automatically discover skills from the configured skills directory and apply them as tasks arise. Start by ensuring the server is reachable through your MCP client, then trigger tasks that align with available skills. For background guidance, you can load an initial skill context at the start of a session to set expectations and workflow. When a task matches a skill, the agent will execute the associated steps and use any referenced resources as needed.
Prerequisites you need before installing: Node.js (recommended latest LTS) and a working MCP client setup.
Choose your platform and paste the following configuration into your project or global MCP config. This example registers a local Skills MCP server using npx and points to a specific skills directory.
{
"mcpServers": {
"skills-mcp": {
"type": "stdio",
"command": "npx",
"args": ["-y", "skills-mcp", "-s", "/absolute/path/to/skills"]
}
}
}You typically configure a single stdio MCP server that runs locally and points to your skills directory. The command and arguments must be exact as shown in examples to ensure the server starts correctly and the agent can discover skills.
If you need to run the server from an environment-specific path, replace the skills directory path with your own absolute path.
When you start the MCP server in your agent, you can load skills at session start using a background guidance prompt. This helps the agent stay aligned with the Skills workflow and ensures it uses the available skills when appropriate.
The server loads skills in a progressive manner and uses metadata, instructions, and resources only as needed to minimize context usage.
Only use skills from trusted sources. A skill can direct agents to execute code or access sensitive data, so validate each skill before enabling it in production.
If you want ready-made examples, you can fetch ready-made skills into your project’s skills directory and then start the server to discover them.
The Skills MCP workflow enables you to discover, load, and use modular skills across agents and platforms, enabling scalable, domain-specific behavior without duplicating knowledge in every conversation.
Lists all available skills with their metadata.
Retrieves the full skill content and absolute path.
Provides informational guidance about the Skills MCP workflow.