🛠️ Skills MCP Server

A centralized, sandboxed Model Context Protocol (MCP) server designed to manage, execute, and expose advanced Agent Skills to any MCP-capable framework or low-code platform (like n8n, custom agents, etc.).

💡 The Problem It Solves

Currently, many popular orchestration frameworks and low-code platforms only support basic tools or MCP. However, modern AI agents require fully encapsulated, complex agent skills (prompts, tool chains, and logic combined). Agno framework natively supports these advanced agent skills, but getting them into platforms like n8n or custom agentic systems has historically been a challenge.

skills_mcp_server is the bridge. By wrapping Agno skills in an isolated FastAPI server and exposing them via the universal Model Context Protocol (MCP), any system that supports MCP can now instantly leverage complex agent skills.

✨ Key Advantages

• 🎯 Selective Skill Exposure (Context Optimization): Install a massive library of skills on the server, but expose only the exact ones a specific agent needs. By tailoring the skill list per agent, you prevent LLM context bloat, save tokens, and drastically improve agent accuracy.
• 🗄️ Centralized Skill Hub: Stop scattering custom scripts and functions across different repositories or n8n nodes. Manage your entire organization's AI capabilities in one unified, easily updatable server.
• 🔒 Sandboxed Execution: All skills run in an isolated environment. You can safely install third-party community skills without risking the integrity of your host application or primary infrastructure.
• 📦 Automatic Dependency Isolation: Skills that include a requirements.txt file are automatically provisioned with a dedicated virtual environment using uv. This ensures perfect isolation and clean uninstalls without bloating the main server's dependencies.
• 🧠 Powered by Agno: Go beyond simple tools. Leverage Agno to build rich, stateful agent skills that can handle complex multi-step reasoning before returning the final payload via MCP.
• 📦 Dynamic Hot-Loading: Add, update, or remove skills on the fly using .zip files, custom .skill packages, or direct download URLs without ever restarting the server.

🏗️ Architecture Concept

The system operates on a dual-interface architecture:

1. REST API (The Manager): A control plane to dynamically install, update, list, and delete skills. It acts as your private skill store.
2. MCP Server (The Provider): The execution plane that securely exposes the curated list of skills to your MCP-capable agents (n8n, Cursor, Claude Desktop, etc.).

🚀 Requirements

• uv (Extremely fast Python package installer and resolver)
• Python 3.12+
• FastAPI
• FastMCP
• Agno

🛠️ Installation & Setup

1. Clone the repository:

   git clone [https://github.com/chameleonbr/skills_mcp_server.git](https://github.com/chameleonbr/skills_mcp_server.git)
   cd skills_mcp_server
   
   

2. Configure environment:

   cp env.example .env
   # Edit .env and set API_KEY and SKILLS_DIR
   

3. Install dependencies:

   uv sync
   

4. Run the server:

   uv run uvicorn main:app --reload
   

Docker

docker-compose up --build

🧪 Running Tests

The project includes a comprehensive unit test suite covering models, routes, services, and the MCP server.

To run the tests, ensure you have the development dependencies installed, and then use pytest:

uv run pytest -v tests/

To run tests with coverage reporting:

uv run pytest --cov=. tests/

🗂️ Project Structure

s

... [View full README on GitHub](https://github.com/chameleonbr/skills_mcp_server#readme)