Detailed Analysis
A reported login redirect loop affecting Claude.ai has surfaced among paying subscribers, with at least one Max x5 tier user documenting the issue via screen recording after standard troubleshooting steps — including disabling ad blockers and cycling through logout/login — failed to resolve it. The bug manifests as an infinite redirect cycle, typically bouncing users between the platform's `/login` and `/onboarding` pages, effectively locking them out of the web interface despite holding an active, paid subscription. The affected user noted the issue persists regardless of ad blocker status, suggesting the root cause lies deeper in the authentication or session management layer rather than in client-side script blocking.
Community research and GitHub issue threads indicate this is not an isolated incident. Multiple users across Anthropic's claude-code repository have flagged similar authentication loops, pointing to a known bug in Claude.ai's login flow. One particularly documented trigger involves the onboarding selection process — specifically choosing "Personal use" during initial or re-triggered onboarding — which can send the session into a perpetual redirect state. Notably, the issue tends to be platform-specific: users experiencing the loop on web or desktop clients have often reported that iOS access continues to function normally, implying the bug is tied to browser-based session handling rather than account-level invalidation.
Recommended workarounds from the broader user community center on browser environment isolation. Opening an incognito or private browsing window — which strips out extensions and starts a clean session state — resolves the loop for many affected users. Clearing cache and cookies or switching to an entirely different browser are secondary options, particularly for cases where the onboarding flow is the suspected trigger. These fixes underscore that the bug likely involves corrupted or conflicting session tokens or cookies persisting from a prior session, which the standard logout flow does not fully purge.
The persistence of this issue among paid subscribers carries meaningful implications for Anthropic's user trust and retention calculus. Max tier subscribers represent Claude.ai's highest-commitment customer segment, and authentication failures that prevent access to a paid product — with no immediate in-product resolution path — create friction that is disproportionately damaging to brand confidence. The fact that the bug has surfaced in tracked GitHub issues without a widely announced fix suggests it falls into an intermittent, hard-to-reproduce category that complicates rapid remediation, a common challenge in OAuth and session-management-dependent web architectures.
More broadly, this type of authentication fragility reflects a systemic growing pain across the AI platform space, where companies like Anthropic are rapidly scaling consumer-facing infrastructure that was originally built for research and API contexts. As Claude.ai competes directly with ChatGPT and Gemini for paid subscriber mindshare, the reliability of core access flows becomes a front-line competitive differentiator. A login loop that forces a paying user to resort to incognito workarounds and Reddit posts — rather than a clear in-app error message and recovery path — highlights the gap between AI capability development and the operational maturity expected of a consumer SaaS product at scale.
Read original article →