RAMbot is an intelligent, autonomous system monitoring and management agent built on the Model Context Protocol (MCP). It pairs a high-performance Python FastAPI backend utilizing psutil with a Node.js agentic loop using LangChain and a local LLM.
Unlike traditional rule-based monitoring tools, RAMbot acts as an autonomous administrator: it evaluates open-ended user inquiries, discovers resource anomalies across all disk partitions, dynamically identifies running processes using an optimized token-matching engine, and takes targeted action to terminate processes when instructed.
RAMbot operates using a decoupled, client-server event-driven architecture connected via Server-Sent Events (SSE):
- Backend (Python Server): Exposes system diagnostic utilities as standardized MCP tools. It interacts directly with the operating system kernel via
psutiland features a globally optimized search matching framework. - Frontend Agent (Node.js Client): Establishes an interactive terminal session, binds the live tool list directly to the LLM's reasoning engine, and drives a continuous evaluation loop (ReAct pattern) allowing multi-step tool chaining.
- Intelligence Layer (Ollama / Qwen3.5): A local LLM highly optimized for complex function calling, tool selection, parameter extraction, and contextual data synthesis.
| Tool Name | Type | Description |
|---|---|---|
get_system_memory |
Read-Only | Returns total capacity, current utilization, and remaining free RAM. |
get_cpu_usage |
Read-Only | Samples CPU execution load across all available logical cores. |
get_disk_usage |
Read-Only | Dynamically scans all mounted storage volumes and physical partitions. |
get_top_processes |
Broad Discovery | Lists the top 3 heavily consuming background application processes. |
find_process |
Target Sniper | Locates running apps via intelligent token-containment analysis. |
kill_process |
Enforcement | Forcefully terminates a specified application thread using its Process ID (PID). |