← work
[systems]2026-03-01

system-eye

AppArmor for macOS resource consumption. Define per-app CPU and memory policies, enforce them automatically. Rust daemon with eBPF telemetry, a CLI, and an MCP layer so AI agents can reason about and adjust system resource allocation.

RustmacOSeBPFCLI

Overview

Most performance tools tell you what's happening — top, Activity Monitor, htop. system-eye goes a step further: it lets you define what should happen and enforces it.

The vision: write a policy like "Xcode gets at most 4 cores and 8GB RAM during a build" or "this background process gets throttled when battery is below 30%" and have the system enforce it automatically, without you babysitting Activity Monitor.

Architecture

A Rust workspace with clean separation of concerns:

systemeye-core       — shared types, pressure classification
systemeye-collector  — system telemetry (CPU, memory, I/O per process)
systemeye-daemon     — background enforcement loop
systemeye-cli        — status, top, why <app>
systemeye-mcp        — localhost HTTP API

The daemon runs continuously, collects telemetry via systemeye-collector, evaluates active policies, and applies enforcement actions (priority adjustment, resource limits via macOS resource controls).

CLI

systemeye-cli status           # current system pressure + active policies
systemeye-cli top --limit 5    # top 5 processes by resource consumption
systemeye-cli why "Xcode"      # why is Xcode being throttled right now

Current status & roadmap

The collector and CLI are working. The daemon loop and policy evaluation engine are in active development.

Next: policy DSL — a simple declarative format for expressing resource constraints:

[[policy]]
app = "Xcode"
max_cpu_cores = 4
max_memory_gb = 8
apply_when = "battery < 50%"

Further out: MCP integration to expose system state and policy management via a local API, enabling AI assistants to reason about and adjust system resource allocation.

This is a tool I want to exist. Building it.