Claude Code network check: methodology & FAQ
Understand what Short ANT can observe, how its network score is calculated, and what a browser cannot verify. The tool runs only when you choose Start check.
Reviewed:
Open network check →What is checked
The full check includes 26 observations across network egress, browser environment, service status, repeated HTTPS requests and WebRTC. Four additional areas are explicitly marked as uncovered. Observation counts are separate from scoring weights.
How the 100-point score works
Only verified evidence earns points. Missing, failed or uncovered evidence earns no confirmed points; covered weights are never expanded to manufacture a perfect score. A low score can reflect missing evidence, not a proven fault.
| Evidence group | Weight | What the points mean |
|---|---|---|
| Transport | 10 | Observable secure connection evidence. |
| Repeated HTTPS egress | 35 | Three same-family rounds with fixed reference and comparison targets. Repeated egress differences do not pass the fixed-exit rule. |
| Repeated WebRTC egress | 35 | Two STUN rounds compared with the same-family HTTPS baseline. Missing public candidates are not proof of safety. |
| Actual DNS path | 20 | Currently uncovered. These 20 points are not awarded. |
Common questions
Can this predict whether my Claude account will be suspended?
No. It observes network behavior, not account enforcement. It cannot determine ban probability, account history, payment status or compliance with service terms.
Does a failed check mean my network is unsafe?
No. A timeout, CORS restriction or blocked endpoint means the evidence was not obtained. The result stays failed or unknown instead of being reported as a pass or a confirmed network fault.
Why do HTTPS targets show different IP addresses?
Routing, proxy rules or service-specific egress can produce different public exits. Only comparable address-family samples are evaluated. Review your authorized network configuration; a browser cannot identify which local process or rule caused a difference.
Does an IPv6 failure prove an IP leak?
No. The browser may have no working IPv6 route or the probe may be blocked. Without a valid response, the check cannot establish an IPv6 exit or infer a leak.
Does WebRTC with no public candidates mean everything is safe?
No. Browser restrictions, network filtering or STUN failure can hide candidates. Short ANT keeps this distinction and does not turn the absence of candidates into an absolute safety conclusion.
Can the page check my Claude Code proxy or firewall?
No. Fixed egress, direct-connection blocking, subprocess inheritance, DNS/UDP/IPv6 rules, TLS and post-update checks are local self-check tips. The webpage cannot inspect your CLI configuration or firewall.
Does normal official status mean my account works?
No. The official service status describes overall service health. It does not verify an individual visitor’s route, supported location, subscription or account access.
What information is collected?
The application does not request Claude passwords, cookies, tokens or API keys. Diagnostic results are masked and kept in page memory; they are not uploaded as a report. Full visitor IP addresses are not stored in the application database or application logs.
Which third parties receive check traffic?
After you start, HTTP probes contact ipify, using ident.me only when the primary source has no valid result; WebRTC contacts Cloudflare STUN. These providers see the connection source IP. Opening this guide does not run probes.
How should I use the result?
Use the evidence to review an authorized network setup, consult the current official service terms and supported regions, and retest after a relevant configuration or software update. Do not treat the score as a guarantee of access or account safety.
Primary sources
Provider policies and service status may change. Check the linked official sources for current information; Short ANT’s scoring rules are its own published methodology.