Know exactly what internet you are getting — and prove it.
Helix runs on your computer and measures your connection without stopping: latency, jitter, loss, outages, speed. When something breaks it tells you where — your Wi-Fi, your router, your ISP, or the far end — and hands you a report you can send to someone who can fix it.
Free tier, no card. Windows 10/11 on x64. The v1.32.1 installer is not code-signed yet — we explain exactly what you will see.
Your ISP measures its network. Nobody measures your line.
ISP-side quality tools are deployed and owned by the operator. Consumer tools are owned by you but tell the ISP nothing. The call ends with “we see no problem on our side” and no way to settle it. Helix is a customer-owned agent whose evidence reaches the customer’s own ISP once that ISP has proved it runs their network — on by default where the network is on the ISP’s own AS number, otherwise once the customer links that ISP and turns sharing on; off with one switch, and never switched back on by anything but the customer — so the conversation starts from the same evidence.
If you just want it to work
A 0–100 score and a plain-language answer to “is it my equipment, my line, or the wider internet?” Outages are logged while you are away from the desk, so the evening the call kept dropping is still on the record tomorrow.
If you do this for a living
Continuous multi-target probing, MTR with per-hop timelines and ASN annotation, loss classification that knows the difference between an ICMP-rate-limited router and real loss, and evidence you can take away: a trace as WinMTR-style text or as an image of every hop, and reports as PDFs or web pages.
If you run the network
A per-subscriber quality view of your customers who run Helix, once you have proved the network is yours: from the day they install if you hold the AS number, otherwise once they link their line and switch sharing on, and each of them can switch it off. Tickets that can arrive with a diagnosis report attached, and an integration contract your NMS can generate types from. The ISP platform →
Four decisions that are unusual, and that we can show you.
Measurement products are easy to make look confident and hard to make honest. These are the four places Helix chose honest, and each one is visible in the product rather than only on this page.
Counts on the wire, never percentages
A percentage cannot be re-aggregated: average four “2 %” windows and you get a number that means nothing. Helix uploads loss counts, the longest consecutive-loss run, and where runs cross a window boundary, so any span you ask for later is recomputed exactly rather than averaged badly.
Below 300 probes, a loss figure is labelled coarse instead of quoted as a rate — because at that sample size it cannot honestly distinguish “fine” from “a bit lossy”.
A verdict has to state what argues against it
Every diagnosis Helix produces carries two lists: supporting evidence and contradicting evidence. Both are always rendered. A verdict engine that cannot populate the second one does not ship.
That rule is what stops the classic false accusation — per-hop loss on a middle router that does not persist to the final hop is a router deprioritising ICMP replies to itself, not a broken path, and the report says so.
Only a verified ISP, and revocation bites immediately
Your figures are shared with your own provider only once it has proved to us that it runs the network your line is on, and never with anyone else. A connection on a verified provider’s own AS number is shared with it by default; on a provider without one, it is shared once you link that provider and turn sharing on. One switch turns either off. Once you turn it off it stays off: nothing automatic ever switches it back on. The next query the ISP runs returns nothing — there is no cache to expire and no nightly job to wait for.
Two independent gates decide it: consent (is sharing on, and still on) and scope (which of your connections this particular ISP may see). Both fail closed. How that works →
The ISP integration contract is public
The schema an ISP’s network-management system integrates against is served unauthenticated, so anyone can read what is exchanged and generate types from it rather than trusting a description.
It is deliberately never stricter than the server: where Helix tolerates a value instead of refusing it, the schema says so. Read the schema →
Install, measure, diagnose, and show your provider.
-
Install the agent
A Windows service does the measuring and a desktop app views it. The split matters: monitoring continues whether or not the window is open, and the app itself never runs elevated.
-
It measures continuously
Multi-target probing at one packet per second, DNS checks every five minutes, bandwidth accounting, scheduled speed tests with a bufferbloat grade. Everything is stored locally first, so the agent is useful before you have an account at all.
-
Run the Link Troubleshooter when something is wrong
A twelve-stage examination (thirteen with the speed test) from your network adapter to the destination — adapter, Wi-Fi radio, gateway, addressing and CGNAT, DNS, captive portal, identity, reachability, path analysis, MTU, nearby content (CDN and transit), throughput, and the last 24 hours of history. Out comes a verdict with its confidence and its counter-evidence, as a report you can save, publish at a revocable link, or attach to a ticket.
-
Show your provider
Your own provider sees your line only once it has proved it runs your network, and one switch, in the app or in your account online, turns that off — nothing but you ever turns it back on. Sign in to your account online for history across devices; from there a support ticket can reach your ISP with the report attached.
Modes change the vocabulary, never the permissions.
The same measurements are worth different things to different people, so the web console has two modes that reorder the panels, change the density and phrase results for the reader. A mode is presentation only — it never grants or withholds a capability, and every panel is present in both.
Home
Is it my equipment, my line, or the wider internet?
Operator
Every number, no rounding and no whitespace.
The other half: how the hosts everyone uses are doing.
Your agent measures your line. NETStats measures the public internet — resolvers, content edges and the servers everything else leans on — from probe boxes we operate in known places on known networks. It is published openly and needs no account, so you can read it before you install anything, and your own agent can show it beside your line once you have.
Subscribers’ lines count toward one thing only: whether a host could be reached at all. The fleet’s figure weighs at half the strength of one of our own boxes, only once at least five different people stand behind it, and never includes a stretch when a subscriber’s own line was down.
Start measuring your own line.
The free tier keeps three days of readable history and costs nothing. The agent runs locally either way.