Connection

Can't reach the backend.

Nothing on this screen can be trusted while the API is down, so every reading has been cleared rather than left showing its last value. {{BRAND}} does not fall back to sample data.

cd backend node server.js
{{BRAND}}PC diagnostics

Sign in with your email and password. Accounts are real now — Argon2id hashing, server-side sessions.

{{BRAND}}PC diagnostics
Marketing site →
—
Health score
—
No scan yet
—

Your machine

—

Waiting for the backend.

Diagnosis

—

—
flagged
—
need you
—
passed

Everything else we checked

—

Live monitor

What the agent is reading right now

Sampled about once a second by the local agent and polled every four seconds by this page. A tile shows — whenever the agent doesn't send that field; it never guesses. Charts are drawn only from samples this browser session has actually received.

Start the live feed

Run this on the PC being monitored

Tiles stay empty until the agent is sampling. Build the exact command below — options change what gets included.

Mode
Network latency target
—

Agent feed

—

Not measured

FPS, frame times and 1% lows — measured, but only while RTSS is running with a game hooked. This agent hooks nothing itself; it reads a block RTSS publishes, so it adds no anti-cheat exposure of its own. Without RTSS these stay blank rather than estimated.

In-game overlay — ruled out permanently. Drawing over a game needs the render-pipeline hook this agent refuses to perform, so all live figures stay in this dashboard.

Per-process background load — nothing reports it, so there is no background-app list.

In-match ping — there's no passive way to discover which server a match is running on. The Network latency tile above is real round-trip time to a fixed reference host (run the agent with --ping), not that.

True DPC/ISR latency — the microsecond-level histogram tools like LatencyMon show needs an admin-rights ETW kernel trace. The DPC/interrupt tile above is a safer, unelevated proxy (% CPU time in DPCs/interrupts), not that measurement.

GPU temperature — needs a vendor SDK thermal query (NVAPI/ADLX); not wired up yet.

Misconfigurations

Every check from the last scan

—

Session history

What you had open, and how the machine coped

Recorded by the agent while Live Monitor runs. Rows are games, not every app you open: a title launched from a Steam, Epic or GOG library is recognised straight away, and a game installed anywhere else is picked up once it holds sustained GPU load for about half a minute. Browsing, chat and desktop work record nothing, and anything under a minute is discarded. The load figures describe the whole machine during the session, not the application itself. Frame rate is included when RTSS was running and supplying frames for that application — rows without it are a record of system load, not of how the game ran.

Benchmarks

Stress tests you've run

A deliberate CPU/GPU load test. Sustain is the figure worth reading: it's how much of its own best throughput the machine was still delivering at the end of the run, so a high score that sustains badly means thermal or power throttling — the thing a raw score hides.

Run one

Runs on the PC being monitored

This pins your CPU and GPU at 100% for the duration — expect fans. It's user-initiated only; nothing schedules it, and the tray app never starts one on its own.

—

Settings

Account & privacy

Everything {{BRAND}} has stored about you, and the two buttons that get rid of it.

Devices

Currently selected: —
Deletes this device and everything recorded for it — scans, sessions and benchmarks. It can't be undone.

Session

Signed in as —

Subscription

Current plan for this device: —
One click, no phone call. Billing is a labelled test stub in this build — no card is ever charged.

What this build does not do

Real authentication, real payments, session recording, benchmarks, FPS capture, GPU temperature and peer percentile comparisons are all unimplemented. Where you see a disabled control, that's why — nothing here is a decorative placeholder pretending to work.

Network latency is measured, but only when the agent is started with --ping <target>, and it is round-trip time to a fixed reference host — not true in-match ping, which nothing here can discover.

Platform · owner console

—

Aggregates over the devices this account actually owns. On a local dev build that is a handful of rows, and the numbers say so rather than borrowing a plausible-looking scale.

Monthly recurring revenue
—
—
Active devices—
Open alerts—
Plan mix—

Devices

—
DeviceHardwarePlanHealthLast seen

Plan mix

Pricing

You spent £2,000 on this rig. This is £6.

Free gets you the diagnosis, forever. Paid keeps the history that turns a one-off reading into a trend — which is the part that actually costs something to keep.

—
Free
One-off check & misconfig scan
—
No card required
  • Full Check Now diagnostic
  • Misconfiguration report with evidence
  • 1 device
  • 48-hour rolling history
Most common
Pro
Live monitoring, one rig
—
Cancel any time from Settings
  • Everything in Free
  • Live Monitor dashboard (no in-game overlay — see below)
  • 30-day history + trend graphs
  • Alerting (Discord, email, push)
  • Your own baseline trending
Creator
CAD / VFX / render workloads
—
Per workstation
  • Everything in Pro
  • 90-day history
  • Viewport & render-engine profiles
  • Scratch disk & VRAM headroom alerts
  • Studio driver validation checks
Fleet
Cybercafes, teams, orgs
—
Minimum 10 devices
  • Everything in Pro, per seat
  • 90-day history + CSV/API export
  • Central admin console
  • Grouped alerting & SLA reporting

Two things this page will not sell you

There is no in-game overlay, on any tier. Drawing over a game means hooking its render pipeline — the exact injection technique anti-cheat systems ban. That's a permanent architectural decision, not a roadmap item, so it isn't listed as a feature you'll get later. There is no peer percentile comparison either until there are enough real machines per hardware bucket to make one honest. Until then you get trending against your own baseline, which needs no one else's data to be true.

Why retainer, not one-time

Nobody blinks at £6/mo to protect a rig that cost thousands, the same way they don't blink at a phone plan. And the value compounds: history, alerting and your own baseline all get better the longer you stay. A one-time licence caps what a Pro user pays in month four, while free stays genuinely useful on its own so it keeps working as the way in.