🚀 Atlas: Agentic MCP Workspace Assistant

Atlas is a professional-grade Model Context Protocol (MCP) ecosystem designed for workspace management and system orchestration. It features a reactive Gradio Dashboard and a high-performance MCP Server that bridges the gap between human users and AI agents.

🌟 Key Features

🖥️ 1. Atlas Dashboard (app.py)

A modern, reactive multi-tab interface:

• 🤖 Smart Assistant: Chat with an LLM that is directly connected to Atlas's local tools.

• 📁 Workspace Explorer: Live view of your project files with real-time status.

• 📈 System Monitor: Real-time gauges for CPU and Memory usage using psutil.

• 📝 Persistent Notes: A structured system to save and retrieve markdown notes.

🏛️ 2. Atlas MCP Server (atlas_server.py)

A comprehensive server that exposes your workspace to local agents (like Claude Desktop):

• 🛠️ Tools:

  • get_weather: Professional real-time weather integration.

  • add_note / list_notes: File-system backed persistent storage.

  • list_files / read_workspace_file: Secure-path-constrained file operations.

• 📚 Resources:

  • system://metrics: JSON-ready system stats.

  • notes://{title}: Dynamic URI scheme for note retrieval.

Related MCP server: mon-serveur-mcp

🏗️ Architecture

graph TD
    User[Human User] -->|Interacts| UI[Gradio Dashboard]
    UI -->|Calls| Server[Atlas MCP Server]
    Agent[Claude/GPT-4o] -->|Connects via stdio| Server
    Server -->|Actions| Files[Local Filesystem]
    Server -->|Monitoring| Sys[System Metrics]
    Server -->|External| Weather[Open-Meteo API]

🚀 Getting Started

1. Prerequisites

• Python 3.10+

• Dependencies: pip install mcp psutil gradio httpx

2. Run the Dashboard

python app.py

This will launch a local server (usually at http://localhost:7860).

3. Connect as an MCP Server

To use these tools inside Claude Desktop, add this to your config:

{
  "mcpServers": {
    "atlas-assistant": {
      "command": "python",
      "args": ["/absolute/path/to/atlas_server.py"]
    }
  }
}

🏛️ Project Structure

• app.py: The reactive frontend.

• atlas_server.py: Core MCP logic & tools.

• .notes/: Directory for persistent workspace data.

• simple_mcp/: Original fundamentals lab.

⚖️ License

MIT License. Created by good2idnan.