A VPN does not solve the core problem with Claude. Anthropic's policy looks at the country where you physically are, not at your IP address, so an account opened through a VPN can be banned at any moment. After a ban you can log in only to appeal, export your data or delete the account, and the export may be limited depending on the violation. If you sit in a country that is not on Anthropic's list (Russia, Belarus, China and Hong Kong are among them), the working replacement for a developer is Claude through an API gateway: no VPN and no Anthropic account. Below are the facts, without the forum horror stories.
Do you need a VPN for Claude?
The question is framed wrongly. A VPN fools an IP check, while Anthropic's Supported Regions Policy describes something else. Its products, from Claude.ai and Claude Code to the API, are available only in the countries and regions on a published list, and anything not on it counts as unsupported. We re-read the policy page on 10 November 2026. Russia is not on it.
The policy names three kinds of unsupported use:
- a person physically located in an unsupported region, even when working for a company from a supported country;
- a company incorporated or headquartered in such a region, in which case the rule covers its staff wherever they sit;
- an organization owned or controlled by people from an unsupported region.
The page carries an effective date of 12 November 2026. The text we read does not contain the word VPN. The consequence is plain: even a perfect VPN does not change where you are, and the policy is about exactly that.
What we know about 12 November 2026 and what the text leaves out
The list has 185 countries and regions, and it does not include Russia, Belarus, China or Hong Kong. The policy also mentions authorized resellers and cloud platforms as channels to Claude models, again only in the listed regions.
The key part is missing. The page does not say how enforcement changes from that date or how the new version differs from the old one, and it does not name independent API gateways. We draw no legal conclusions and do not promise that access through any service will stay as it is.
What is known about VPN-linked bans
There is less confirmed here than the headlines suggest. We sorted the claims by source.
| Claim | Status |
|---|---|
| Russia and Hong Kong are not on the supported list | Confirmed by Anthropic's policy |
| Account creation from an unsupported location is named as a ground for a ban | Confirmed by Anthropic's help article on safeguards, warnings and appeals |
| Affected users get emails about “suspicious signals” and a Usage Policy violation, with an appeal offered | Reported by Kod.ru on 2 October 2026 |
| A VPN was the cause | A suspicion voiced by users and the outlet. Anthropic has not confirmed it publicly |
| Appeals do not help | User reports: some say the decision was called final |
| Scale of the wave | No official figures |
Kod.ru, a Russian tech outlet, also writes that an earlier wave hit users in Russia in May, with the same suspected causes: frequent IP changes and VPN use. We pass these stories on as user reports. They do not become facts until Anthropic confirms them.
Why a VPN does not fix it
The first reason sits in the policy itself: it is tied to physical location, and a VPN does not move you. The second is payment. A card from a bank in a restricted region is usually declined whatever your IP says, and Paying for Claude in restricted regions shows what works and what does not. The third is the price of a mistake. A ban ends normal use of the whole account, and the data export that remains may be limited depending on the violation.
There is a fourth reason. A third-party VPN service controls your route and sees network metadata: the addresses you connect to and how much traffic you send. HTTPS keeps the content of requests, keys included, unreadable to it unless something intercepts or breaks the encryption, but you still hand your connection to its operator. For that reason we do not rank VPNs for Claude and do not name any. We have no scheme for getting around a regional restriction and we will not publish one.
What to do if your account was banned
The order below follows Anthropic's help article on warnings and appeals.
- Read the email and find the stated reason. The article lists repeated Usage Policy violations, account creation from an unsupported location and Terms of Service violations.
- File an appeal. Log in to the banned account at claude.ai and fill in the form at
claude.ai/restricted. Without a login the form is unreachable. Anthropic names no reply time and says it is answering more slowly than usual. - Take your data. Free, Pro, Max and Ultra users can log in to a banned account, export their data or delete it. The export may be limited depending on the violation.
- Do not open a new account the same way. The same location gives the same ground for a ban.
- Move your working setup to an API key, so that losing an account does not stop the work.
If you got a suspension notice rather than a ban, read “Your account has been suspended” in Claude: what to do. The fuller recovery plan is in Claude Account Banned, Suspended, or Deleted.
How to stop depending on one account
Keeping a whole workflow inside someone else's account is risky, because one email can end it. Make a few habits mandatory.
- Keep agent instructions in your repository.
CLAUDE.md, skills and prompt templates belong in git, not as the single copy inside Projects. - Export the chat history you would hate to lose, on a schedule.
- Never tie scripts and bots to a personal subscription. Automation gets its own API key, which you can revoke and replace.
- Write code so that changing the model provider means editing one address and one variable.
The last habit is easier than it sounds. Through Claudexia, models from different vendors are called with one key, and the OpenAI-compatible address https://api.claudexia.tech/v1 works with most SDKs.
Claude without a VPN: how it works through the API
Claudexia does not restore accounts and does not open claude.ai. It is a gateway to model APIs: you get a key that starts with sk_cdx_, top up the balance with a card, SBP or crypto, and call Claude models at https://api.claudexia.tech. No VPN is involved, and you never create an Anthropic account.
Claude Code connects through environment variables. Keep the ANTHROPIC_DEFAULT_* lines: without them the agent asks for its own default models and gets an error. A permanent setup through settings.json is described in the Claude Code guide.
export ANTHROPIC_BASE_URL="https://api.claudexia.tech"
export ANTHROPIC_AUTH_TOKEN="sk_cdx_YOUR_KEY"
export ANTHROPIC_MODEL="claude-sonnet-5"
export ANTHROPIC_DEFAULT_OPUS_MODEL="claude-opus-5-5"
export ANTHROPIC_DEFAULT_SONNET_MODEL="claude-sonnet-5"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="claude-sonnet-5"
claude
A direct call to the Messages API looks like this:
curl https://api.claudexia.tech/v1/messages \
-H "x-api-key: sk_cdx_YOUR_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "content-type: application/json" \
-d '{
"model": "claude-sonnet-5",
"max_tokens": 256,
"messages": [{"role": "user", "content": "Hello"}]
}'
Prices are in USD per 1M tokens, from the Claudexia Premium price list of 10 November 2026; current figures are on the models page. Claude Opus 5.5 costs $0.86 for input and $3.30 for output, Claude Sonnet 5.5 costs $0.52 and $1.98, and Claude Haiku 5.5 costs $0.10 and $0.36.
What you do not get: the claude.ai site and the account features of Anthropic, such as Projects, memory across conversations and the Claude in Chrome extension. Those belong to the subscription, not to the API. The Claude Desktop app can connect to the gateway, and the setup is in the Claude Desktop guide.
One caveat for companies. Anthropic's policy explicitly mentions organizations registered in unsupported regions. Claudexia is a separate service with its own terms, and we give no legal opinion on how a vendor's regional rules apply to your company. If this matters to you, talk to a lawyer before you build a product on it.
How to tell an honest service from a trap
Many offers float around the search for “Claude without VPN”, and some are dangerous. Be wary if you are offered any of the following:
- a shared Claude Pro or Max account for two or three people for a symbolic fee: the owner risks a ban, and so do you;
- a request to hand over the login and password of your own account “to set up the subscription”;
- an executable or a “patch” that is not on the official download page;
- an “unlimited Claude” with no contract, no description of terms and no way to check which model you are actually talking to.
Working API access is duller. It has a public model list with IDs and prices, plain addresses, a key you create and revoke yourself, and a balance that is spent per token. The Claudexia model list is open at GET https://api.claudexia.tech/v1/models, and How to Use Claude API Without an Anthropic Account explains how the access is built.
What else to use when you have no account
| Task | Route |
|---|---|
| Agent in the terminal | Claude Code on a gateway key, setup in the Claude Code guide |
| App with MCP | Claude Desktop in Gateway mode |
| Bots, CI, your own services | API billed per token, paid in rubles or crypto |
| Other models | GPT, Gemini, DeepSeek, Kimi and Qwen through the same key |
| Anthropic subscription features | Claudexia does not sell subscriptions, and availability depends on Anthropic's regions |
If Claude is not a must, compare the options in Claude & Claude Code Alternatives in 2026. What the Pro and Max subscriptions cost against the API is in Claude Pro, Max or API: which is cheaper.
FAQ
Do I need a VPN for Claude?
Not if you work through an API gateway. Claude.ai and an Anthropic account fall under Anthropic's regional rules, and a VPN does not change them, because the policy is about where you physically are. API access through Claudexia needs no VPN.
Does Claude ban for using a VPN?
Anthropic's official documents do not mention VPNs. The help article names account creation from an unsupported location as a ground for a ban. Users and media tie the waves of bans in Russia and Hong Kong to VPNs, and Anthropic has not publicly confirmed that.
How can I use Claude without a VPN?
Through the API. Get a Claudexia key, top up the balance with a card, SBP or crypto, and set https://api.claudexia.tech in Claude Code, Claude Desktop or your own code. You need no Anthropic account, but you get no claude.ai chat and none of the subscription features.
Which VPN is best for Claude?
We do not name services and do not offer workarounds. A VPN does not change the physical location the policy refers to, and a third-party service controls your route and sees network metadata such as addresses and traffic volume. Removing the dependency on one account and moving to the API is safer.
What changes on 12 November 2026?
The Supported Regions Policy page shows an effective date of 12 November 2026. Russia is not on the list, and the text calls use by people physically located in an unsupported region, and by organizations registered or controlled there, unsupported. The page does not explain how enforcement will change after that date.
Does Claude Code work without a VPN?
Yes, if you point it at a compatible gateway: set ANTHROPIC_BASE_URL, a key and the default models. Install Claude Code from the official addresses and connect the key as described in the Claude Code guide. Signing in to an Anthropic account is not needed in this mode.
What if Claude banned my account?
Log in to the banned account at claude.ai, fill in the appeal form at claude.ai/restricted and export your data. Expect to wait: Anthropic says replies are slower than usual. In parallel, move your work to the API so that you do not stop.