skill4agent

skill4agent’s MCP server provides a lightweight way to search, view details, and install AI skills within conversations. It exposes three primary tools via the MCP protocol: search_skills to discover skills by keyword with optional category filters; get_skill to fetch detailed documentation for a selected skill (including SKILL.md); and install_skill to obtain installation commands or download links for a given skill. The server can be run locally or integrated into client apps (e.g., Claude Desktop) by declaring the skill4agent MCP server in the mcpServers configuration and, if needed, supplying a custom API endpoint via the SKILL4AGENT_API_URL environment variable.

To use it, configure your MCP client to point to the skill4agent server entry shown in the example. Then call search_skills with a keyword to find relevant skills, use get_skill with a skillId from the results to read full details, and call install_skill with the same skillId to retrieve installation instructions. You can also override the default API URL by setting SKILL4AGENT_API_URL in env to a custom endpoint that your infrastructure exposes.

Prerequisites:

• Node.js and npm installed on your system
• Internet access to fetch the MCP server package from npm

Installation steps (global):

1. Install the MCP server globally via npm:

npm install -g @skill4agent/mcp-server

2. Run the MCP server directly with npx (no global install required):

npx @skill4agent/mcp-server

Usage in a client configuration (example):

1. Add the MCP server to your client configuration as shown:

{
  "mcpServers": {
    "skill4agent": {
      "command": "npx",
      "args": ["-y", "@skill4agent/mcp-server"]
    }
  }
}

2. If you want to point the server to a custom API endpoint, set the environment variable:

{
  "mcpServers": {
    "skill4agent": {
      "command": "npx",
      "args": ["-y", "@skill4agent/mcp-server"],
      "env": {
        "SKILL4AGENT_API_URL": "https://your-custom-domain.com/api/mcp"
      }
    }
  }
}

Note: The SKILL4AGENT_API_URL can be used to direct the MCP requests to a custom backend if needed.

Tips and notes:

• The server integrates three tools: search_skills, get_skill, and install_skill. Use them by passing the appropriate parameters (e.g., keyword, skillId).
• By default, the API endpoint is https://skill4agent.com/api/mcp. Override it with SKILL4AGENT_API_URL if your deployment uses a different domain.
• If you run into port or network issues, ensure your environment allows outbound HTTP(S) requests and that any proxy configurations are respected.
• For development, you can run npm run dev or build and execute node dist/index.js as described in the Local Development section of the repo.
• The MCP server is designed to be integrated into larger AI assistant workflows; ensure your client respects the MCP schema when sending requests and handling responses.