Checks your network as Claude sees it: a supported region, no proxy or datacenter exit, one exit for all Anthropic domains, and no mainland-China traits.
We look at your connection the way Claude does: the exit IP it sees and whether that region is supported, whether the IP is flagged as a VPN, proxy or datacenter, whether your real IP leaks through WebRTC or DNS, and whether your local network and system contradict the exit. Takes about 5–10 seconds.
Exit IP profile
Supported regions
VPN detection & real IP leaks
Local network
Local system
No sign-in. Environment traits are computed locally; IP intelligence and leak probes need the network. History stays on this device
Claude region and IP requirements at a glance
Whether Claude works for you comes down to three things: first, is the exit IP Claude sees in one of Anthropic's 185 officially supported countries and regions (mainland China, Hong Kong and Macau are not; territories of listed countries are supported too); second, is that IP flagged as a VPN, proxy, datacenter or Cloudflare WARP; third, do claude.ai, the Anthropic API and the Console use the same exit, does WebRTC leak your real IP, and do your time zone, language and similar settings carry mainland China traits. Phone number, payment method and your company's ownership are account-level and can't be checked from outside.
Anthropic officially supports 185 countries and regions, and Claude.ai, Claude Code and the API share one list; territories and dependencies of listed countries are supported too (checked on 2026-10-21).
Not supported: mainland China, Hong Kong, Macau, Russia, Belarus, Iran, North Korea, Syria, Cuba, Venezuela, Afghanistan, Myanmar, Yemen and others; the occupied Ukrainian regions of Crimea, Donetsk, Kherson, Luhansk and Zaporizhzhia are excluded as well.
Since September 2025, companies controlled from unsupported regions (including entities more than 50% owned, directly or indirectly, by such companies) may not use Claude either, regardless of their exit IP.
claude.ai, api.anthropic.com and platform.claude.com all sit behind Cloudflare; their traces show the IP and country Anthropic really sees.
Claude Code uses the exit and system time zone of the machine it runs on; when it runs on a server, in WSL or in a container, check that machine.
A DNS leak doesn't change the address Anthropic sees (queries only reach resolvers); a WebRTC leak hands the real IP behind your proxy to any page script.
Based on Anthropic's supported countries page (Claude.ai and the API share one list): 185 countries and regions are available. Mainland China, Hong Kong, Macao and sanctioned countries are not.
Any country or region not on the supported list counts as unsupported; the official page states that territories and dependencies of listed countries are also supported (e.g. Puerto Rico, Guam, Gibraltar). The occupied Ukrainian regions of Crimea, Donetsk, Kherson, Luhansk and Zaporizhzhia are also unsupported, but IP location stops at country level so this page can't tell them apart. The official list changes — the official page is authoritative.
Claude region & IP FAQ
Q1
Can I use Claude in mainland China or Hong Kong?
No. Mainland China, Hong Kong and Macau are not on Anthropic's official list (185 supported countries and regions), and access from IPs there is refused. Taiwan, Singapore, Japan, South Korea and the United States are supported.
Q2
Are Claude bans related to my IP?
Region is a hard requirement. Within supported regions Anthropic hasn't published its risk rules, but shared VPNs, proxy pools, datacenter IPs, frequent country hopping and real-IP leaks all raise the odds of verification and review in industry practice. This page scores each of these common signals and shows the evidence for each.
Q3
Do Claude Code and the claude.ai web app have the same IP requirements?
The region rules are the same: one shared list. The difference is IP type and where the exit is: running Claude Code with an API key on a server is common, while web sign-up and sign-in are more sensitive to datacenter IPs. Claude Code uses the exit and time zone of the machine it runs on; check that machine with the curl command on this page.
Q4
Will using a VPN with Claude always get me banned?
Not necessarily. The problem usually isn't the VPN itself but whether the exit is shared by many users, flagged by commercial databases, split across Anthropic's domains, or leaking your real IP through WebRTC. A dedicated, fixed exit in a supported region is far lower risk.
Q5
Why does this site see a different IP from Claude?
Your proxy app sends different sites through different exits (split routing). This page goes by claude.ai's own trace; split routing itself costs no points, only different exits across Anthropic's domains, a WebRTC leak of another IP, or IPv4 / IPv6 in different countries do.
Q6
Is Claude supported in territories like Puerto Rico, Guam or Gibraltar?
Yes. Anthropic's official page says territories and dependencies of listed countries are also supported, so IPs in US, UK, French and other overseas territories follow their parent country. Hong Kong and Macau are not such cases.
Q7
If the check says low risk, is my account safe?
No. Your phone number's country, payment card billing address, your company's ownership and your account history live inside Anthropic and can't be checked from outside. A clean environment only rules out network-layer problems.
Q8
Does this check log my IP?
IP intelligence lookups go through our server and create ordinary access logs; we don't sell or share lookup records. Browser traits such as time zone, language and fonts are computed locally on your device and never uploaded; exit IP history stays in this browser only.
About the mainland-China trait check
Community reverse-engineering of Claude Code claims that when ANTHROPIC_BASE_URL points to a third-party relay, Claude Code reads the local system time zone, combines it with the relay hostname to judge whether you are in mainland China, and writes that into the system prompt sent to the model. This claim comes from the open-source project FuckClaude; Anthropic has not confirmed it and CleanIP has not independently reproduced it.
A web page cannot read everything Claude Code sees on its machine, but it can read the same system time zone. On top of that this tool weighs browser language, Intl locale, Chinese fonts not shipped with the OS, Chinese vendor fonts, Chinese browsers and Chinese devices — 7 signals in total; bands: low 0–30, medium 31–60, high 61–100. Signals with no discriminating power — Chinese fonts preinstalled on every Mac / Windows, flag emoji that no Windows machine renders — are not scored.
Web: any script on claude.ai can read the browser time zone and language; whether Anthropic uses them for risk control is not public. Claude Code: community analysis says it reads the system time zone when connecting to unofficial endpoints. Note that Claude Code reads the time zone of the machine it runs on — use the terminal commands above to check.
No. Of the 7 signals only "system time zone" corresponds to the mechanism described by the community; the rest are closely related Chinese-environment traits. The score is an estimate, not a verdict, and it is only part of the "Browser environment" item in the overall score.
Keep the system time zone consistent with the exit IP location and move zh-CN off the top of the browser language list; the machine running Claude Code needs the same adjustment (e.g. set the TZ environment variable).
No. All 7 environment traits are computed locally in your browser and never uploaded.
Privacy note
All 7 environment traits are computed locally in your browser and never uploaded. The other checks on this page need the network: looking up exit IP intelligence, reading exits from Anthropic domains, contacting public STUN servers for WebRTC, and the DNS leak probe.
The official supported-regions page lists countries/regions for Claude.ai and the commercial API separately; the two lists are currently identical at 185 entries, and mainland China, Hong Kong and Macau are not included. This page judges region by the exit Claude sees.
If the exit is located in an unsupported region, Claude cannot be used normally even without any VPN, proxy or abuse record.
Ukraine's Crimea, Donetsk, Kherson, Luhansk and Zaporizhzhia regions are also unsupported, but IP geolocation usually stops at country level, so this page cannot tell them apart.
Geolocation providers can disagree; this page prefers Cloudflare's location (Claude sits behind Cloudflare) and cross-checks it with CleanIP's own.
Company ownership is restricted too (account level — not checkable here)
Since September 2025, Anthropic prohibits companies or organizations whose ownership structure subjects them to control from unsupported regions, wherever they operate — including entities more than 50% owned, directly or indirectly, by companies headquartered in unsupported regions.
This restriction is about who the user is and has nothing to do with the exit IP.
Phone number country, card billing address and organization details are also account-level and cannot be checked here.
The score on this page only answers whether the network and browser environment exposed by this visit has problems.
Policy Violations Can Trigger Restriction, Suspension, or Termination
Anthropic's transparency page says that if a user violates its policies it may take enforcement actions such as warning, suspending or terminating access; banned users can request a review.
Anthropic does not publish its risk rules; industry anti-fraud systems commonly weigh account, region, access environment and security signals together.
Shared proxies, frequent cross-border jumps, abuse history and high-risk networks all raise review likelihood in industry practice.
The VPN / proxy / Tor, abuse and exit-consistency items on this page map to these common signals.
Enterprise VPN and Anonymous VPN Are Different Signals
Claude Enterprise's IP allowlisting documentation shows that organizations can submit CIDRs for offices, VPN egress points, and other approved access points, so organizational access comes only from trusted network locations.
An enterprise VPN is a controlled, auditable, registered egress; anonymous VPNs or proxy pools usually lack this context.
If a company enables an allowlist, unregistered IPs are denied access — missing a required CIDR can lock users out of the organization.
So a "VPN hit" doesn't necessarily mean a violation, but shared VPNs, proxy pools, and unstable origins significantly raise the risk.
Claude's active sessions page shows device, browser, IP-based approximate location, and last-updated time, helping users spot unfamiliar sessions and terminate suspicious logins.
An account that frequently shows up from unfamiliar countries, unfamiliar devices, or unusual browser combinations is more likely to be flagged for account security anomalies.
IP location, system timezone, browser language, and WebGL/UA fingerprint should stay as consistent as possible.
If an unrecognized location appears in active sessions, the official guidance is to terminate that session and consider changing your password.
