Monty watches your systems so you don't have to.
A local-first daemon that alerts you before problems become crises.
Monty is the Monitor Lizard — a systems monitoring daemon and ratatui TUI that watches your machines, evaluates configurable alert rules, and notifies you when something is about to go wrong.
Unlike hosted monitoring SaaS, Monty runs entirely on your hardware. No accounts, no cloud, no egress. It reads your systems, applies your rules, and speaks — literally, if you want — when thresholds are crossed.
┌─────────────────────────────────────────────────────────┐
│ ● daemon:ok ⚠ 2 alert(s) 12:34:07 │
├─────────────────────────────────────────────────────────┤
│ [Alerts ⚠2] [Metrics] [History] [Rules] │
├─────────────────────────────────────────────────────────┤
│ ACTIVE ALERTS │
│ ● CRIT nuc disk./.used_pct 96% firing 5m │
│ ⚠ WARN gnuc cpu.percent 92% firing 2m │
├─────────────────────────────────────────────────────────┤
│ gnuc CPU ████████░░ 82% MEM █████░ 64% DSK ██░ 42%│
│ nuc CPU ██░░░░░░░ 21% MEM ███░░░ 48% DSK ████ 96%│
├─────────────────────────────────────────────────────────┤
│ daemon:ok collectors:4 q:quit 1-4:tabs ↑↓:scroll │
└─────────────────────────────────────────────────────────┘
Because problems don't announce themselves. Disks fill silently. CPU spikes happen at 3 AM. Memory leaks grow for days before something crashes. By the time you notice, the damage is done — a job failed, a service dropped requests, a deadline was missed.
Monty breaks that pattern. It watches continuously, evaluates rules you define, and interrupts you before the threshold becomes a disaster:
- Disk at 90%? Voice alert. Terminal bell. Before it hits 100% and corrupts data mid-write.
- CPU pinned at 95% for 5 minutes? Notification. Before the build queue backs up and engineers start filing tickets.
- Remote machine stopped heartbeating? Alert. Before your users notice the service is down.
The difference between Monty and a dashboard is who initiates contact. A dashboard waits for you to look. Monty comes to find you.
| Monty | Grafana Alerting | PagerDuty | cron + scripts | |
|---|---|---|---|---|
| Local-first, no cloud | ✓ | ✗ | ✗ | ✓ |
| Zero config to start | ✓ | ✗ | ✗ | ✗ |
| Voice alerts | ✓ | ✗ | ✗ | maybe |
| Prometheus integration | ✓ | ✓ | ✓ | ✗ |
| NATS / swarm-aware | ✓ | ✗ | ✗ | ✗ |
| SSH-polled remotes | ✓ | ✗ | ✗ | ✓ |
| Ratatui live dashboard | ✓ | browser | browser | ✗ |
| Single static binary | ✓ | ✗ | ✗ | ✓ |
# Build and install to ~/bin
just install
# Run the live dashboard with default config (monitors localhost)
monitor-agent tui
# Check collectors and voice engine
monitor-agent doctor
# Print active alerts as JSON
monitor-agent alertsZero config needed to get started. By default Monty monitors the local machine with three built-in rules: CPU > 85%, disk > 90%, memory > 90%.
Monty pulls metrics from wherever your systems live:
| Collector | How it works | Refresh |
|---|---|---|
| Local | sysinfo crate + nvidia-smi subprocess |
2s |
| Prometheus | HTTP /api/v1/query against any Prometheus endpoint |
30s |
| SSH | Runs one command over SSH, parses /proc + df |
30s |
| NATS | Subscribes to subjects; parses JSON metric payloads | real-time |
Mix and match. A typical homelab config monitors the local machine via
sysinfo, remote machines via Prometheus node_exporter, and the
gilabot agent swarm via NATS heartbeats — all at once.
Rules live in monitor-agent.toml:
[[rules]]
name = "high-cpu"
target = "*" # "*" matches all targets
metric = "cpu.percent"
condition = { gt = 85.0 }
severity = "warn"
cooldown_secs = 300
message = "{target}: CPU at {value:.0}%"
[[rules]]
name = "critical-disk"
target = "*"
metric = "disk.used_pct"
condition = { gt = 90.0 }
severity = "critical"
cooldown_secs = 3600Conditions: { gt = N }, { lt = N }, { eq = N }
Severities: info, warn, critical
Cooldown: minimum seconds between repeated alerts for the same rule + target
When a rule fires, Monty dispatches to every configured channel:
| Channel | What happens |
|---|---|
| Terminal bell | \x07 BEL + printed alert line — works in any terminal |
| Voice | Auto-detects say (macOS), espeak-ng, piper, or PowerShell TTS |
| NATS publish | Publishes a JSON alert event to a configurable subject |
| Webhook | HTTP POST to any endpoint (Slack, Discord, custom handler) |
Configure in monitor-agent.toml:
[notify]
terminal_bell = true
voice = true
voice_engine = "auto" # auto-detects best available engine
nats_subject = "monitor.alerts"
webhook = "" # set to a URL to enable[daemon]
socket = "" # default: $XDG_RUNTIME_DIR/monitor-agent.sock
[[targets]]
name = "local"
kind = "local"
[[targets]]
name = "gnuc"
kind = "prometheus"
endpoint = "http://192.168.0.104:9090"
[[targets]]
name = "nuc"
kind = "ssh"
host = "192.168.0.104"
user = "hartsock"
key = "~/.ssh/id_ed25519"
[nats]
servers = ["nats://192.168.0.104:4222"]
subjects = ["swarm.heartbeat", "monitor.>"]
[[rules]]
name = "high-cpu"
target = "*"
metric = "cpu.percent"
condition = { gt = 85.0 }
severity = "warn"
message = "{target}: CPU at {value:.0}%"
[[rules]]
name = "critical-disk"
target = "*"
metric = "disk.used_pct"
condition = { gt = 90.0 }
severity = "critical"
[notify]
terminal_bell = true
voice = true
voice_engine = "auto"
nats_subject = "monitor.alerts"Config is searched in order: MONITOR_CONFIG env → ./monitor-agent.toml
→ ~/.config/monitor-agent/config.toml → /etc/monitor-agent/config.toml
→ built-in defaults.
# Prerequisites: Rust 1.80+, just
cargo install just
# Full quality gate (fmt + clippy + test + 80% coverage floor)
just check
LLVM_COV=/opt/homebrew/opt/llvm/bin/llvm-cov \
LLVM_PROFDATA=/opt/homebrew/opt/llvm/bin/llvm-profdata \
just cov-ci
# Install release binary to ~/bin
just install
# Install pre-push hook (enforces the same gate before every push)
just install-hooksmonitor-agent/
├── monitor-core/ Types, config, alert engine, Collector/Dispatcher traits
├── monitor-collect/ Collectors: local, prometheus, nats, ssh
├── monitor-alert/ Dispatchers: terminal bell, voice, webhook, nats-pub
├── monitor-tui/ ratatui dashboard (Alerts, Metrics, History, Rules tabs)
└── monitor-cli/ Binary entry point — daemon, tui, status, alerts, doctor
Next: Phase 11 — IPC socket daemon/TUI split (daemon runs as a background
service; TUI attaches on demand). See docs/ROADMAP.md.
Built with Rust · Powered by ratatui · Mascot by the Gilamonster Foundation