Claude OAuth Failures: Claude has started having serious OAuth errors
Anthropic fixed Claude Code OAuth outages after Cloudflare 403 errors and hanging authorization spinners disconnected developers across Europe and Asia.
By Dillip Chowdary • Nov 14, 2026 • Source: HN Claude/Codex/Fable
In this post
Anthropic experienced widespread authentication failures across its Claude platform and Claude Code utility, locking engineers out of active terminal workflows and generating Cloudflare gateway rejections. The breakdown caused sudden account logouts, broken credential renewal redirects, and recurring verification failures, as documented in HN Claude/Codex/Fable's report. Users reported that login attempts stalled with infinite browser spinners, while Downdetector recorded rapid surges in service incident reports.
This report details the technical sequence of the authentication breakdown, the Cloudflare denial messages encountered by engineers, and the temporary workarounds observed during terminal sessions. It is written for software developers and engineering teams who rely on automated AI command-line tooling and need verified details on how connection state disruptions impact running operations.
01Cloudflare 403 blocks and authentication loops
Developers attempting to execute commands inside Claude Code encountered abrupt terminal halts accompanied by API Error 403 Attention Required notices from Cloudflare. The interface prompted engineers to run the login command, only for credential exchanges to cycle through short operational bursts before failing again. Terminal sessions recorded cycles where authentication held for twenty seconds at 11:15 before crashing into Cloudflare errors at 11:16.
The failure pattern produced severe instability for active subscribers who cycled through repeated login completions only to face immediate rejection. PeterStuer documented a sequence where Claude Code churned for nineteen seconds, completed login successfully, and collapsed twenty-two seconds later with another Cloudflare error. The cyclical lockout occurred independently of user account tier, striking paid accounts holding active subscriptions.
Even when client terminals displayed recurring Cloudflare warnings, underlying language model operations occasionally continued generating responses between authentication crashes. User Sx999 observed that Claude Code continued answering prompts and executing tool calls sporadically across two hours despite generating authentication failure warnings. The condition left the interface alternating unpredictably between functional query execution and terminal access blocks.
02Browser redirect hangs and authorization spinners
The secondary failure mode emerged within the web browser authorization handoff used to validate local client sessions. When engineers followed terminal prompts to initiate credential refreshes, the application opened browser sessions pointed at the claude.com authentication path containing cai oauth authorize parameters. Rather than presenting the expected permission grant dialog or returning a token, the page hung indefinitely displaying a loading spinner.

The browser hang prevented users from completing clean credential handshakes even after multiple manual refreshes. User joquarky reported that the browser window opened the authorization link directly but sat permanently on the spinner graphic without completing the handshake. Similar behavior was documented by dibbla, who established an initial connection but found subsequent browser validation attempts stalled on the identical loading animation.
Terminal attempts that failed to clear the authorization gate returned explicit OAuth request failed warnings to the command-line interface. User abendrothj confirmed that direct authentication requests threw explicit failure exceptions when attempting initial handshakes. While some engineers like tankenmate eventually succeeded by repeatedly retrying the authorization link, the underlying redirect pathway remained prone to unpredictable timeouts.
Advertisement
Tech Pulse Daily
Get tomorrow's pulse first
Join engineers who read Tech Pulse before stand-up. Free, weekday mornings.
03Mid-task session drops and terminal interruptions
The disruption hit long-running autonomous operations particularly hard, terminating background developer scripts without saving session momentum. User joquarky documented an abrupt Cloudflare 403 authentication crash that severed an active background process after running uninterrupted for twenty-seven minutes and ten seconds. The interruption demonstrated that session authentication tokens expired or invalidated in mid-flight rather than waiting for command boundaries to conclude.
Similar multi-stage failures disrupted engineers managing persistent code-generation tasks across regular working routines. In Cracow, Poland, developer mac_ek reported being forcibly logged out five distinct times across a single thirty-minute window, primarily immediately following intensive background jobs. Although repeated attempts occasionally restored access, session validity collapsed again after fifteen minutes of sustained operation.
Engineers discovered an unexpected escape route inside the Claude Code interface by utilizing client cancel commands. Developer MarcBuilds01 observed that pressing cancel when presented with an authentication failure bypassed the hard exit and returned the user to the active prompt for several turns. This temporary client behavior allowed programmers to preserve running session context despite the active failure of the OAuth service.
04Global developer disconnects across multiple regions
The authentication outage exhibited worldwide scope, impacting programming teams across European, Asian, and Mediterranean network regions simultaneously. In Graz, Austria, developer A_D_E_P_T was ejected from Claude Code and locked out for twenty minutes before regaining momentary access. Neighboring engineer franz_matt reported initial recovery in Graz, though subsequent connection attempts across the area collapsed again after five minutes of stability.
Regional reporting rapidly confirmed that the service interruption stretched far beyond Central European infrastructure. Developer aj6718 reported identical disconnects in India where sessions functioned for brief minutes before forcing sudden logouts, while Miffles201912 experienced connection drops in Malaysia. Additional corroborating outage reports were filed simultaneously by developers located across Germany, Italy, Poland, and the Netherlands.
Public monitoring infrastructure captured the widespread nature of the breakdown as individual reports multiplied on developer forums. Monitoring dashboard Downdetector showed sharp error spikes for Claude services as programmers found themselves unable to sustain coding sessions. Commenter fisher_sun noted the complete stall of automated programming workflows, while developer vvolhejn logged brief five-minute recoveries that collapsed back into systemic lockouts.
05Status portal updates and incident mitigation
During the opening stages of the failure, Anthropic's public operational dashboard showed no active alerts for general authentication or chat services. User tankenmate pointed out that status.claude.com initially omitted any acknowledgment of the widespread authentication breakdown during the first several minutes of user lockouts. The only degraded status listed on the dashboard was an earlier ongoing advisory concerning the Microsoft Office 365 connector.
Anthropic subsequently acknowledged the severity of the outage by publishing official incident report nn11wk485zfn to its public tracking infrastructure. User joquarky clarified that the widespread login crashes were entirely distinct from the pre-existing Microsoft Office connector trouble, as affected developer workloads did not interface with Office software. The published incident report provided official confirmation as connection stability gradually improved across affected regions.
Recovery occurred unevenly as authentication tokens slowly regained validity across regional edge networks. Approximately two hours after the onset of sporadic Cloudflare errors, user Sx999 confirmed that session connections stabilized completely without requiring a manual re-login command. As the official incident reached resolution, engineers successfully refreshed their authorization URLs to resume standard automated development without encountering further 403 challenge pages.
Claude OAuth Failures FAQ
What error message appeared during the Claude Code authentication failures?
Users encountered API Error 403 Attention Required notices from Cloudflare along with OAuth request failed alerts during login attempts.
How did the OAuth authorization page behave in the browser?
The authorization link opened a URL containing cai oauth authorize parameters on claude.com but remained stuck on an infinite loading spinner.
Which regions reported authentication drops and session disconnections?
Developers confirmed outages across Austria, Poland, Germany, Italy, India, Malaysia, and the Netherlands.
Did Anthropic publish an official incident report for the disruption?
Anthropic posted official incident report nn11wk485zfn to its status portal at status.claude.com as connection stability began recovering.
Sources
Author
Dillip Chowdary
Writes Tech Bytes coverage of AI, engineering, and the tools that actually ship. Editor of Tech Pulse Daily.
Related on Tech Bytes
Advertisement