- 0third parties — every request goes to this server only
- 0permission prompts — nothing asks for camera, location or notifications
- 0stored — the report exists in your browser and nowhere else
What this is
A live answer to “what does a website learn about me just by being visited?” — the same information any site could gather, laid out with its source, its reliability and its meaning. It measures rather than dramatises: guesses are labelled as guesses, and nothing is inflated to look alarming.
How it works
- The server layer. To serve any page at all, this server necessarily receives your public IP address, request headers and connection details. Those rows are marked server.
- The browser layer. The page’s own JavaScript reads what your browser exposes to every website automatically — screen, language, hardware hints, graphics, fonts, audio and more — and runs a few same-origin tests (canvas rendering, a bait script for content blockers, a DNS lookup that only our own nameserver answers). Those rows are marked js.
- The verdicts. The Overview combines rows into answers: is this a VPN or relay, where does your DNS go, how automated does the client look, what fingerprint does it leave. Each verdict lists the signals it used.
Finding your way around
- Overview — the verdict, your address and four facts at a glance.
- Leaks — the four checks a VPN should pass: DNS, the wire, a port on your address, the path back to you.
- Fingerprint — the hash, a private-window comparison, how an ad system would file you, a safe-to-share image.
- Speed — round trip, throughput, bufferbloat, cold DNS and compute, against this server.
- Report — every observation, searchable, with copy and download.
- Terminal — the same answers as plain text for
curl and scripts.
Reading a row
- Direct — shown exactly as received. Derived — computed from direct values. Inferred — an educated guess with a stated confidence (browser family from the User-Agent, for example).
- Reliability says how far to trust it: exact, browser-reported, coarsened by the browser, estimated, or transient.
- Masked values (cookies, full URLs, credential-like headers) stay hidden until you reveal them, and are excluded from copies and exports unless you tick “include sensitive values”.
- Rows the browser withholds or does not support are filed under each section’s technical details, with the reason.
The leak checks
- VPN / relay verdict. Compares the network your IP belongs to (consumer ISP, hosting, CDN, mobile carrier, a published iCloud Private Relay egress range), whether your browser’s time zone agrees with the IP’s, and connection characteristics. It is a weighed judgement with a confidence, never a claim of certainty.
- DNS leak test. Your browser resolves a random name under
t.wget.me, a zone served only by this server’s own nameserver. Whoever asks for that name is your resolver — your ISP, a public resolver, or your VPN’s. If your traffic leaves through a VPN but the lookup came from your ISP, that is a leak, and it says so.
- Port check and traceroute. The port check and “The path back to you” connect or probe towards your own address only. They cannot be pointed at anyone else.
- The wire. The first packets of your connection reveal the operating system that sent them (TTL, window, option order) and the TLS stack that started the handshake (the JA4 fingerprint). Compared with what your User-Agent claims, they expose spoofed agents, VPN re-origination and TLS-intercepting proxies.
Where the data comes from
Location and network-operator rows use database files stored on this server — MaxMind GeoLite2 (refreshed twice a week) with DB-IP Lite as a fallback — plus Apple’s published list of Private Relay egress ranges. Lookups never leave the server; no external IP service is ever queried about you. The footer names the database in use. IP geolocation is city-level at best and can be wrong, which is why those rows say “approximate”.
What this page never does
- No analytics, no tracking pixels, no CDN, no fingerprinting vendor, no external lookups.
- No permission prompts. A section lists what a page could ask for, and that it did not.
- No cookies, no local storage, no service worker — nothing is written to your browser.
- No copy of the report is stored, logged or transmitted. What you copy or download stays with you.
What a web page cannot read
Despite what some sites imply, JavaScript on a page cannot read your MAC address, IMEI or serial numbers, your Wi-Fi password, arbitrary files, saved passwords, other sites’ cookies or storage, your browsing history, other tabs, your name, or a complete list of installed apps and extensions. Where browsers leak indirect clues, this report labels them as inference, never as fact.
What is logged
The web server keeps standard access logs (source IP address, request time, URL, status, browser User-Agent and referrer). They are rotated daily and retained. The inspector application itself stores nothing. The DNS test nameserver keeps a short rolling log of the random names it answered and the resolver addresses that asked — that is how the leak test works. A passive observer on this server keeps a similar rolling log of the first packets of connections to this site only (the TCP handshake and the plaintext TLS ClientHello — nothing is decrypted), so the “wire” rows can be shown back to the connection they belong to. Neither log identifies you beyond the connection you are looking at, and both truncate themselves.
About fingerprints
The hashes in the Fingerprinting section summarise measurements taken during this visit. Another site running similar tests could recognise a similar browser configuration, which is why they are shown. Without a population to compare against, no result can be called “unique”, so this page does not.
From a terminal
Everything here is also available as plain text: curl wget.me prints your IP, /check the VPN verdict, /dns the leak test. See the full reference — output formats are frozen, so scripts keep working.