A community perspective on account stability
The following is an external author’s perspective, not verified documentation of Anthropic’s systems.
Anthropic appears to treat account trust as a moving score rather than a simple on/off switch. That score is shaped by registration, payment, usage, device stability, and network behavior. Understanding those inputs explains why some accounts survive heavy use while others are blocked after a few days.
1. Registration sets the baseline
The first signals come from the IP, email provider, and the client used to create the account. A clean residential IP and a trusted mailbox create a stronger starting position than a disposable email or a heavily used proxy.
- Clean IP: the most important factor.
- Mailbox: some providers carry more weight than Gmail.
- Immediate KYC: accounts that require document checks immediately are less stable.
- Client: mobile registration appears to add trust.
2. Payment behavior changes the score
Subscription changes are interpreted as risk events. An immediate Max upgrade can look like a sudden shift in value, while a slower progression from Pro to Max appears more natural.
- Timing: subscribing immediately after registration can lower the score.
- Plan: start with Pro, then move to Max after a pause.
- Card: a real foreign card is better than a virtual one.
- KYC: a document request during payment signals observation, not an automatic ban.
3. Usage patterns reveal automation
Heavy parallelism, rapid limit exhaustion, and constant re-login look more like automation than normal human work. Even useful activity can lower trust when it happens too fast.
- Do not exhaust the monthly limit in one sitting.
- Prefer desktop and mobile apps over constant web access.
- Use the memory feature to make long-term usage look intentional.
- Avoid parallel sessions when possible.
4. IP and device stability matter more than perfection
Anti-fraud systems are more sensitive to sudden change than to a specific IP. A stable residential IP, one region, and one client produce a more believable pattern than a “perfect” proxy that changes every few minutes.
- Keep one IP, one region, and one client.
- Use global proxy mode with TUN enabled.
- Avoid frequent IP rotation and multi-device logins.
- Reduce high-parallel operations and repeated re-login attempts.
5. Account age adds resilience
Older accounts have a larger trust reserve. Accounts created in 2023 and earlier tend to survive more adverse events. The score grows slowly, but it compounds over time.
- Old accounts are noticeably more stable.
- Score increases gradually with normal use.
- Major plan changes are safer on mature accounts.
6. Observation and audit are separate stages
When an account approaches the block threshold, Anthropic may place it in an observation period. A typical sign is a login reset. During this stage, additional risky behavior can push the account over the limit.
- Observation: login resets are a common warning sign.
- Audit: suspected ToS violations can trigger a home-directory scan by Claude Code.
- Artifacts: exploits, PoCs, and hacking tools in the home folder increase ban risk.
What ClaudeBrew changes
ClaudeBrew focuses on the network part of this equation: residential IP stability, consistent TLS and headers, careful token handling, and a reverse proxy for document verification. It does not fake documents or “hack” Claude; it makes the environment look like a normal, stable workstation.
Source: analysis on X