macOS Resource Optimizer

Production-ready system optimization with 40+ specialized agents for comprehensive macOS resource management.

Quick Reference

What is macOS Resource Optimizer? Real-world macOS optimization framework with 40+ specialized agents executing in parallel:

coordinator.py: 40-agent orchestrator (6 phases, 4-5s execution)
40+ specialized agents: Memory, disk, browser, Docker, developer tools
Implementation: UV scripts (PEP 723) + Bash delegation via MoAI agents

Main Orchestrator:

Script	Purpose	Agents	Execution Time
`coordinator.py`	40-agent parallel orchestrator	40 agents (6 phases)	4-5s

6 Phases (coordinator.py):

Disk Cleanup (15 agents): Python/Node zombies, Browser helpers, Network leaks, Docker containers
RAM Optimization (9 agents): Memory pressure, App profiler, Browser tabs, Electron apps
Developer Cache (5 agents): Time Machine, Xcode, Build caches, Docker cleanup
Advanced Memory (4 agents): Swap optimizer, WindowServer, Spotlight, Memory leaks
Browser Deep Cleanup (3 agents): Chrome, Safari, Firefox optimizers
App & System (3 agents): Messaging apps, VSCode, DNS/Network

Performance:

Sequential: 40 × 1.0s = 40s (estimated per agent)
Parallel (6 phases): 4-5s total (8× faster than sequential)
Real-world: 4-7s depending on system state and cache availability
With MetricsCache (TTL 30s): ~2-3s on repeated calls

Usage

1. Full System Optimization (40 agents)

# Execute all 40 agents in 6 parallel phases
uv run scripts/coordinator.py

# JSON output
uv run scripts/coordinator.py --json

2. Individual Agents

# Memory pressure detector
uv run scripts/agent_memory_pressure_detector.py

# Browser tab manager
uv run scripts/agent_browser_tab_manager.py

# Docker cleanup
uv run scripts/agent_docker_deep_cleanup.py --dry-run

3. Utility Scripts

# Kill zombie processes
uv run scripts/kill_zombies_parallel.py

# Report memory usage
uv run scripts/report_memory.py

# Analyze running processes
uv run scripts/analyze_processes.py --json

MoAI Integration

Manager Agents

manager-resource-coordinator.md:

# Execute full 40-agent orchestration
result = Bash("uv run .claude/skills/macos-resource-optimizer/scripts/coordinator.py --json")
data = json.loads(result.stdout)

# Parse results by phase
phase1_results = data["phases"]["disk_cleanup"]
phase2_results = data["phases"]["ram_optimization"]

# Return aggregated recommendations

Expert Agents

expert-memory-optimizer.md:

# Execute memory-specific agents
result = Bash("uv run scripts/agent_memory_pressure_detector.py --json")
memory_data = json.loads(result.stdout)

# Generate recommendations based on memory analysis

Available Agents (40+)

Phase 1: Disk Cleanup (15 agents)

Process Cleanup:

agentpythonzombies.py - Python zombie processes
agentnodeprocess_scanner.py - Node/Bun zombie processes
agentworkerdzombies.py - Cloudflare Workers zombies
agentgenericidle.py - Generic idle process hunter
agentjvmmemoryhogdetector.py - JVM memory hog detection
agentsshgitprocesszombies.py - SSH/Git process zombies

Application Helpers:

agentbrowserhelpers.py - Chrome/Arc renderer helpers
agentlanguageservers.py - VS Code language servers
agentelectronhelpers.py - Notion/Dia helpers

Network & Resources:

agentnetworkconnection_leaks.py - Network connection leaks
agentorphanedprocess_groups.py - Orphaned process groups
agentdockercontainer_scanner.py - Docker container scanning
agentdatabaseconnection_pooler.py - Database connection pooling
agentsshconnection_scanner.py - SSH connection scanning
agentfilecache_optimizer.py - File cache optimization

Phase 2: RAM Optimization (9 agents)

agentmemorypressure_detector.py - Memory pressure analysis
agentbrowsertab_manager.py - Browser tab management
agentbrowserhelper_consolidator.py - Browser helper consolidation
agentbrowsercache_optimizer.py - Browser cache optimization
agentinactiveapp_detector.py - Inactive application detection
agentelectronapp_optimizer.py - Electron app optimization
agentbackgroundapp_suspender.py - Background app suspension
agentswapoptimizer.py - Swap usage optimization
agentmemoryleak_hunter.py - Memory leak detection

Phase 3: Developer Cache (5 agents)

agenttimemachinesnapshot_cleaner.py - Time Machine snapshots
agentdevelopercache_cleaner.py - Developer cache cleanup
agentxcodecache_cleaner.py - Xcode artifact cleanup
agentbuildcache_cleaner.py - Gradle/Maven cache cleanup
agentsystemlog_cleaner.py - System log cleanup

Phase 4: Advanced Memory (4 agents)

agentswappurgeable_hunter.py - Purgeable swap memory
agentwindowserver_optimizer.py - WindowServer optimization
agentspotlightmds_hunter.py - Spotlight MDS optimization
agentmemoryleak_hunter.py - Memory leak detection

Phase 5: Browser Deep Cleanup (3 agents)

agentchromedeep_cleanup.py - Chrome deep cleanup
agentsafarioptimizer.py - Safari optimization
agentfirefoxdeep_cleanup.py - Firefox cleanup

Phase 6: App & System (3 agents)

agentmessagingapp_hunter.py - Messaging app optimization (Slack/Discord)
agentvscodedeep_cleanup.py - VS Code cleanup
agentdnsconnection_scanner.py - DNS/Network optimization

Architecture

Execution Stack

User Command (slash command)
    ↓
MoAI Command (Python orchestrator)
    ↓
Task() delegation to manager agents
    ↓
Manager-Resource-Coordinator (MoAI agent)
    ↓
Bash(uv run coordinator.py) → UV Script execution
    ↓
asyncio.gather() parallel execution
    ├─ Phase 1: Disk Cleanup (15 agents)
    ├─ Phase 2: RAM Optimization (9 agents)
    ├─ Phase 3: Developer Cache (5 agents)
    ├─ Phase 4: Advanced Memory (4 agents)
    ├─ Phase 5: Browser Cleanup (3 agents)
    └─ Phase 6: App & System (3 agents)
    ↓
JSON results aggregation
    ↓
User-facing report (Korean)

Implementation Details

Execution Method: UV Scripts (PEP 723)

#!/usr/bin/env uv run
# /// script
# requires-python = ">=3.11"
# dependencies = ["psutil", "pyyaml"]
# ///

import asyncio
import psutil

# Scripts run directly via: uv run script.py
# No Python virtual environment setup required

Delegation Pattern: Bash + Task()

# Manager agent receives command
# Delegates to Bash tool: uv run .claude/skills/.../scripts/coordinator.py
# Coordinator spawns async tasks for 40 agents
# Results aggregated and returned

Data Flow

# coordinator.py executes agents
{
    "phases": {
        "disk_cleanup": {
            "agents_executed": 15,
            "duration": 2.1,
            "savings_gb": 5.3,
            "results": [...]
        },
        "ram_optimization": {
            "agents_executed": 9,
            "duration": 1.8,
            "memory_freed_gb": 2.1,
            "results": [...]
        },
        ...
    },
    "summary": {
        "total_agents": 40,
        "total_duration": 2.5,
        "total_savings_gb": 12.4,
        "total_memory_freed_gb": 4.2
    }
}

Protected Apps

Default protected apps (from config/cleanup-rules.json):

Claude Code
Notion
Slack
Discord
Mail
Messages
Ghostty

Recommended additional protection (for development environments):

Node.js (active development processes)
Apple Virtualization (system virtualization)
VSCode/Cursor (development editors)
Xcode (development tools)
Docker Desktop (containerization)

Customization: Edit config/cleanup-rules.json to add/remove protected apps based on your workflow.

These apps are NEVER killed or suspended during optimization.

Performance Characteristics

Metric	Value
Total Agents	40+ specialized agents
Orchestrators	1 (coordinator only)
Execution Time (parallel)	4-5s (first run), 2-3s (cached)
Execution Time (sequential)	~40s (estimated)
Speed Improvement	8× faster (parallel vs sequential)
Memory Saved (typical)	1-3 GB
Disk Saved (typical)	0.4-2.5 GB
Actual Results (2025-11-30)	+413MB disk, 18% of goal

Commands Integration