This is a local, intelligent monitoring server built for the Model Context Protocol (MCP). It provides deep insight into your system and running processes, enabling a local LLM agent to analyze the machine’s state, perform diagnostics, suggest optimizations, and offer real-time solutions.
Note
This project was tested on the latest versions of macOS. While most features work reliably on this version, compatibility with other operating systems (Linux/Windows) is not always guaranteed.
This project provides tools to:
- 🔧 Retrieve a full system summary (OS, CPU, RAM, Disk, Boot Time)
- 📊 Monitor CPU usage in real time
- 💾 Track RAM and swap memory usage
- 💽 Check disk usage (total, used, free, % used)
- 🧩 Analyze top processes sorted by CPU load
- 🔎 Search for running processes by name
- 🌲 Explore the process tree of a given PID
⚠️ Detect anomalies like:- High memory + low CPU (e.g., memory leaks)
- Zombie/defunct processes
- Resource usage spikes above a threshold
- 🧠 Deliver actionable warnings and insights
This allows a local agent to not only observe, but also understand what’s happening on your machine — and provide real-time, specific solutions based on the actual system state.
It is recommended using uv, so make sure you have uv installed. Generally, for more information regarding MCP checkout MCP Python SDK.
Clone this repo then:
uv syncWith uv:
uv run mcp install main.pyOr directly:
mcp install main.pyTo test in development mode with logging:
mcp dev main.pyNote
🧠 Note: When using Claude Desktop, ensure the app is properly restarted after installing a new server.
Find more information on MCP Quickstart
The project lays the foundation for a powerful local system intelligence layer. In future versions, one could:
- Expand full cross-platform support
- Add background process logging
- Integrate energy usage and thermal monitoring
- Allow custom alert thresholds and scheduled diagnostics
- Let agents make automated corrective actions