Network

Pip probes two targets every 2 seconds: your router, and an internet anchor beyond it. That pair is enough to localise a fault, so the diagnosis is one sentence in plain words — from “Your connection looks healthy” through “Your local network is fine — the problem starts past your router” to “The problem starts at home: your Wi-Fi, cable, or router.”

Under it: latency as a live chart, with jitter and loss beside it. A lost probe draws a gap in the chart, never a zero — a dead moment should look dead, not fast. The probes use the same unprivileged API as ping itself, so none of this needs administrator rights; where ICMP is filtered, Pip falls back to a TCP measurement and labels it as one.

On demand only — it moves a few hundred megabytes, so it never runs itself. You get download and upload, idle latency, latency while the line is busy, and a bufferbloat grade from A+ to F. That last one is the number behind “my internet is fast but calls stutter”: if latency climbs several times over under load, the line is quick and still unpleasant to use.

Tests run against the nearest of Pip's own nodes — Sydney, Singapore, Los Angeles, Frankfurt — and the payloads are built to be incompressible, so nothing between you and the node can flatter the number.

Run diagnostics

Benchmarks four DNS resolvers, checks IPv4 and IPv6 separately, then runs every reading Pip already holds through a set of rules. Each finding is a title, an explanation, and a probable cause — not a wall of numbers.

Host check

Type a host — api.example.com, or db.internal:5432 — and get DNS, TCP over v4 and v6, and HTTPS, each timed, with a verdict sentence. It is the answer to “is it me, or is it them?”

Looking for listening ports and bind addresses? That lives with the server it belongs to — each server's Ports tab shows every address it is bound to, and warns when one is reachable from your whole network.