Every host. One live table.
Concurrent ICMP and TCP ping monitor for maintenance windows, on one fixed latency scale, with JSON evidence at the end.
uv tool install pinghue
| host | address | state | last | min | avg | max | jitter | loss | mode | history |
|---|---|---|---|---|---|---|---|---|---|---|
| edge-router-1 | 192.168.1.1 | healthy | 14.79 | 11.64 | 14.30 | 18.43 | 1.01 | 0.00% | icmp | |
| core-sw-1 | 10.0.0.10 | healthy | 10.75 | 10.50 | 12.01 | 14.27 | 0.81 | 0.00% | icmp | |
| db-primary | 10.0.1.20 | intermittent | 449.31 | 20.11 | 57.96 | 449.31 | 17.53 | 2.26% | icmp | |
| api-gw | 10.0.2.30 | healthy | 36.76 | 34.00 | 39.38 | 46.85 | 2.59 | 0.00% | icmp | |
| backup-nas | 192.168.50.10 | down | 8.31 | 7.05 | 9.35 | 13.15 | 1.35 | 15.25% | icmp | |
| dns-resolver | 8.8.8.8 | healthy | 14.92 | 10.62 | 13.26 | 17.19 | 1.29 | 0.00% | icmp |
TUI when you watch, lines when you script
One command, three ways to run it. Every mode probes your hosts concurrently and keeps the same whole-run statistics: by default ICMP gives every host its own probe slot (up to 5,000 hosts), and TCP runs up to 1024 probes at a time with --concurrency.
| host | address | state | last | min | avg | max | jitter | loss | mode | history |
|---|---|---|---|---|---|---|---|---|---|---|
| 1.1.1.1 | 1.1.1.1 | healthy | 4.08 | 3.23 | 3.67 | 4.36 | 0.23 | 0.00% | icmp | |
| 8.8.8.8 | 8.8.8.8 | healthy | 3.78 | 2.83 | 3.51 | 4.22 | 0.43 | 0.00% | icmp | |
| 9.9.9.9 | 9.9.9.9 | healthy | 10.62 | 9.71 | 10.24 | 11.08 | 0.31 | 0.00% | icmp |
The default. Concurrent echo probes with the full TUI: history bars, states, and keyboard control. Press a to show or hide the resolved addresses.
| host | address | state | last | min | avg | max | jitter | loss | mode | history |
|---|---|---|---|---|---|---|---|---|---|---|
| 1.1.1.1 | 1.1.1.1 | healthy | 6.42 | 5.88 | 6.31 | 7.02 | 0.19 | 0.00% | tcp | |
| 10.0.4.12 | 10.0.4.12 | down | - | - | - | - | - | 100.00% | tcp |
No privileges needed. Connect checks against any port from 1 to 65535. A refused connection prints its own amber ! in the history, so you can tell a closed port from a timeout at a glance.
$ pinghue -p 443 example.com -c 3 --no-tui
2026-10-06T02:31:04.118204+00:00 example.com ok latency=14.08ms
2026-10-06T02:31:05.119927+00:00 example.com ok latency=13.92ms
2026-10-06T02:31:06.121376+00:00 example.com ok latency=14.31ms
$ echo $?
0
For cron and CI. Line output is the default when stdout is not a terminal; on a terminal, pass --no-tui. One line per probe, exit code 3 when a --fail-on-any-down or --fail-on-all-down condition triggers, and --output for the report.
One fixed scale, every run
Latency maps to the same eight glyphs on every run. Nothing rescales, so rows stay comparable across hosts and across nights. Green until 100 ms, amber past it.
- up to 1 ms
- up to 3 ms
- up to 10 ms
- up to 30 ms
- up to 100 ms
- up to 300 ms
- up to 1000 ms
- over 1000 ms
- lost, down
- TCP refused
A host that blips once stays intermittent until you reset it with r. The report covers the whole window, not only its final minutes.
Close the window with a report
Add --output and pinghue writes a schema-versioned JSON summary when the run ends. Attach it to the change ticket and go back to bed.
- Versioned contract
schema_version: 1. Breaking changes need a new schema version.- Whole-run stats
statscovers every probe in the run;samplesis the recent tail, at most 1000 per host.- Private by default
- Reports are written
0600, and an existing file is never replaced without--overwrite. - Exit codes
0completed,1runtime error,2usage error,3a fail-on-down condition triggered.
$ pinghue -f hosts.txt --duration 180 \
--output maintenance.json
{
"schema_version": 1,
"pinghue_version": "8.0.1",
"run": {
"started_at": "2026-10-06T02:28:08.412Z",
"ended_at": "2026-10-06T02:31:08.418Z",
"host": "local",
"exit_reason": "deadline",
"samples_window": 1000,
"probe": {
"mode": "icmp", "port": null,
"interval_s": 1.0, …
}
},
"targets": [
{
"target": "db-primary",
"resolved_address": "10.0.1.20",
"resolved_family": "ipv4",
"status": "intermittent",
"error": null,
"stats": {
"sent": 180, "received": 176,
"loss_pct": 2.22,
"min_ms": 20.11, "avg_ms": 64.07,
"max_ms": 451.44, "jitter_ms": 23.59
},
"samples": [ … ]
},
…
]
}
Small on purpose
A focused local tool for maintenance windows, migrations, and quick reachability checks. Nothing more.
- Not a Prometheus, Smokeping, Zabbix, or NMS replacement.
- Not a long-running metrics database or alerting system.
- Not a privileged daemon.
- Not a packet capture or traceroute tool.
- Not a service that accepts remote network requests.
Sixty seconds to a live table
uv tool install pinghuepipx install pinghuebrew install inxbit/tap/pinghuepython -m pip install pinghuemacOS and Linux, Python 3.11 to 3.14. Run pinghue --check first: on Linux, ICMP may need a ping_group_range setting, and TCP mode (-p) needs none.